advertiment. lʼaccés als continguts dʼaquesta tesi queda ... · mentre que la regió perianal i...
Post on 11-Aug-2019
218 Views
Preview:
TRANSCRIPT
ADVERTIMENT. Lʼaccés als continguts dʼaquesta tesi queda condicionat a lʼacceptació de les condicions dʼúsestablertes per la següent llicència Creative Commons: http://cat.creativecommons.org/?page_id=184
ADVERTENCIA. El acceso a los contenidos de esta tesis queda condicionado a la aceptación de las condiciones de usoestablecidas por la siguiente licencia Creative Commons: http://es.creativecommons.org/blog/licencias/
WARNING. The access to the contents of this doctoral thesis it is limited to the acceptance of the use conditions setby the following Creative Commons license: https://creativecommons.org/licenses/?lang=en
Variability and diversity in the
skin of healthy dogs:
microbiome, genome and environment
Anna Maria Cuscó Martí
Departament de Ciència Animal i dels Aliments
Facultat de Veterinària, Universitat Autònoma de Barcelona
A thesis submitted for the degree of Doctor (PhD)
Director: Dr. Olga Francino Martí
Tutor: Dr. Armand Sánchez Bonastre
June 2017
La Dra. Olga Francino Martí, investigadora del Departament de Ciència Animal i dels
Aliments de la Universitat Autònoma de Barcelona
Certifica que Anna Maria Cuscó Martí ha dut a terme sota la meva direcció el treball de
recerca realitzat al Departament de Ciència Animal i dels Aliments de la Facultat de
Veterinària de la Universitat Autònoma de Barcelona, que ha portat a l'elaboració d'aquesta
Tesi Doctoral, titulada «Variability and diversity in the skin of healthy dogs: microbiome,
genome and environment».
Bellaterra, 2 de juny de 2017
Dra. Olga Francino Martí Anna Maria Cuscó Martí
This work was supported by a grant awarded
by Pla de Doctorats Industrial (2013 DI
011) provided by the Agencia de Gestió
d’Ajuts Universitaris i de Recerca (AGAUR);
Secretaria d’Universitats i Recerca del
Departament d’Economia i Coneixement de
la Generalitat de Catalunya
This Industrial PhD project was performed
in collaboration with Vetgenomics, SL
Anna Maria Cuscó Martí was funded by
Vetgenomics, SL.
i
Resum
En la darrera dècada han incrementat molt els estudis del microbioma –conjunt de microbis
que habiten un microhàbitat comú– gràcies a la millora de les tècniques de seqüenciació
massiva. Aquesta nova manera d’analitzar les comunitats microbianes ha permès detectar una
gran diversitat abans ignorada en infinitat d’ambients diferents, entre ells diferents parts del
cos.
El microbioma associat a certes parts del cos presenta un gran interès per la seva interacció
amb les cèl·lules i la immunitat de l’hoste. En un estat normal, aquestes interaccions
microbioma – hoste promouen la salut de l’individu. En canvi, en una patologia, aquestes
interaccions es troben alterades.
L’objectiu d’aquesta tesis és definir la variabilitat pròpia del gos sa, principalment a nivell de
microbioma cutani però també en immunitat innata. Per tal d’aconseguir aquest objectiu
utilitzem tècniques de seqüenciació massiva i procedim a: i) caracteritzar la immunitat innata
dels gossos sans a nivell de polimorfisme en Toll-like Receptors –primers sensors de patògens; ii)
caracteritzar el microbioma de la pell en dues cohorts de gossos sans; iii) provar la
seqüenciació de tercera generació (single-molecule seqüencing) per nanopors aplicada a la
caracterització del microbioma.
S’ha caracteritzat la diversitat genètica del Toll-like Receptors en una població d’estudi amplia per
representar la gran diversitat genètica pròpia de les espècies de cànids (set races diferents de
gossos i dues poblacions de llops). S’han descobert noves variants genètiques, i s’han detectat
d’altres prèviament identificades. S’ha dissenyat i validat un xip de genotipatge individual amb
64 sondes per detectar algunes de les variants amb possible efecte a la proteïna (mutacions no
sinònimes), ja que són les que poden estar afectant a la resposta immunitària.
El microbioma de la pell en gossos sans s’ha avaluat en 8 regions diferents de la pell i en dues
cohorts de gossos sans. S’han inclòs diferents regions representatives de la diversitat cutània i
se n’han pogut determinar certes característiques comunes. Entre totes les regions de la pell
incloses, la cara interna del pavelló auricular és la que presenta valors més grans de diversitat
mentre que la regió perianal i nasal les que presenten valors més baixos. Tot i això, hem
detectat que les diferents regions cutànies són més semblants en un mateix individu, tan en
gossos d’ambient diferents com en gossos vivint en un mateix espai i interaccionant entre ells.
Per tant, l’efecte individual és el factor més determinant tan de l’estructura com de composició
del microbioma de la pell dels gossos. Hem d’entendre aquest efecte individual, com un
conjunt entre l’hoste, el seu ambient i el seu comportament.
Finalment, hem provat una nova estratègia per caracteritzar el microbioma basada en la
seqüenciació de tercera generació per nanopors. L’estratègia més usual en l’estudi de
microbioma està basada en la seqüenciació d’un gen marcador de bacteris i arqueobacteris
(16S rRNA) utilitzant fragments curts i seqüenciadors de segona generació. La seqüenciació de
ii
tercera generació (single-molecule seqüencing) permet seqüenciar fragments més llargs d’ADN,
permeten analitzar tot el gen marcador sencer. Tot i la baixa precisió de la tècnica escollida,
hem detectat encara més diversitat en la pell dels gossos i hem pogut assignar la taxonomia a
nivells inferiors. La tecnologia avaluada té molt de potencial, però probablement altres
aproximacions experimentals donarien resultats més satisfactoris en l’estudi del microbioma
mitjançant gens marcadors.
En conclusió, s’han pogut ampliar els coneixements en la variabilitat pròpia del gos sa sobretot
a nivell de microbioma de la pell, però també en immunitat innata. Per altra banda, aquest
projecte de Doctorat Industrial conjunt amb Vetgenomics ha permès obrir i posar a punt una
nova línia estratègica a l’empresa.
iii
Summary
The community of microorganisms that inhabits a specific environment is named as the
microbiome. In the last decade, microbiome studies have boosted due to the apparition and
development of next-generation sequencing (NGS) techniques that allow massive sequencing
of DNA. This new way of analyzing microbial communities allowed detecting a wide range of
diversity previously ignored in many different environments, including the body.
The microbiome associated to different body sites, interacts to host cells and immunity to
contribute to the host functions as well as promote health. In contrast, in a pathology these
microbiome-host interactions are altered.
The aim of this thesis is to define the intrinsic variability of the healthy dog, mainly at skin
microbiome level, but also at innate immunity. To achieve this goal we use massive sequencing
techniques and proceed to: i) characterize the innate immunity of healthy dogs at
polymorphism level in Toll-like Receptors, which are the first sensors of pathogens; ii)
characterize the skin microbiome of two cohorts of healthy dogs; iii) test the third generation
sequencing (single-molecule sequencing) by nanopores applied to microbiome studies.
We characterized the genetic diversity of the Toll-like receptors in a wide population to
represent the genetic diversity of the canid species (seven different breeds of dogs and two
populations of wolves). New genetic variants have been discovered and others previously
identified have also been detected. We designed and validated a genotyping chip with 64
individual probes to detect some of the variants with a likely effect in the protein (non-
synonymous mutations), since they may be affecting the immune response.
The skin microbiome in healthy dogs has been evaluated in eight different regions of the skin
and in two cohorts of healthy dogs. Different regions representing the skin diversity have
been included and some common trends identified. Among all the skin regions, the inner part
of the ear presented higher alpha diversity values, whereas nasal and perianal regions
presented the lowest ones. However, we found that different skin regions on the same dog
resemble more than the same skin region in different dogs, either in dogs living in different
environments or in dogs living in the same space and interacting with each other. Therefore,
the individual effect is the main factor driving the structure and composition of the skin
microbiome in dogs. We must understand this effect as the individual, together with its
environment and its behavior.
Finally, we tested a new strategy to characterize the microbiome based on third generation
sequencing by nanopores. The usual approach in microbiome studies is targeting of a marker
gene for bacteria and archaea (16S rRNA) with short primers and second-generation
sequencers. The third-generation sequencing (single-molecule sequencing) can sequence long
DNA fragments, and hence analyze the full-length marker gene (1,500 bp). Despite the low
accuracy of the technique chosen, we were able to detect even more diversity in the skin of
dogs and we could assign sequences to lower taxonomic levels. The technology assessed has a
iv
great potential, but probably other experimental approaches would give better results in the
study of the microbiome by marker genes.
In conclusion, we have been able to expand the knowledge on the variability of the healthy
dog, especially at the skin microbiome level, but also at innate immunity one. Moreover, this
Industrial Doctorate project together in collaboration with Vetgenomics allowed us to open
and consolidate a new strategic business area for the company.
v
Contents
1. Introduction .................................................................................................................................... 1
1.1. Conducting a microbiota study ............................................................................................ 2
1.1.1. Experimental considerations ........................................................................................ 5
1.1.2. Next generation sequencing (NGS) ............................................................................ 9
1.1.3. Bioinformatics analysis ................................................................................................ 14
Alpha diversity ....................................................................................................................................... 18
Beta diversity.......................................................................................................................................... 19
1.2. Overview of the skin............................................................................................................ 23
1.2.1. Anatomy and physiology of the canine skin ............................................................ 24
1.2.2. Histology of the canine skin ....................................................................................... 25
1.2.3. Skin immunity ............................................................................................................... 27
1.3. Skin microbiota ..................................................................................................................... 35
1.3.1. What is living on healthy skin? ................................................................................... 36
1.3.2. Factors shaping bacterial skin microbiota ................................................................ 41
1.3.3. Commensal microbiota functions on the skin ......................................................... 43
1.3.4. Skin microbiota and dermatological diseases ........................................................... 46
1.3.5. Skin microbiota as a clinical tool ............................................................................... 48
2. Objectives ..................................................................................................................................... 53
3. Results ............................................................................................................................................ 55
3.1. Paper 1 pagina de titol com a portada ............................................................................... 56
3.2. Paper2 pagina de titol com a portada ................................................................................ 56
3.3. Individual signatures and environmental factors shape skin microbiota on healthy
dogs 82
4. Discussion ................................................................................................................................... 131
4.1. Dual assessment of innate immunity and skin microbiota........................................... 133
4.2. Skin site signatures on canine skin microbiota .............................................................. 137
4.3. Individual signatures on skin microbiota ........................................................................ 139
4.4. Environmental bacteria on skin: transient or resident microbiota? ............................ 142
4.5. Conducting microbiota studies: 16S rRNA gene and beyond ..................................... 144
5. Conclusions................................................................................................................................. 151
vi
List of Boxes
Box 1. Summary of the main nomenclature used on microbiota studies. ...................................... 4
Box 2. Types of natural selection. ...................................................................................................... 31
List of Tables
Table 1. Main sequencing platforms used in microbiota studies and their main characteristics.
................................................................................................................................................................. 14
Table 2. Dermatological diseases associated with an alteration on Toll-like Receptors. ............ 34
Table 3. Propionibacterium acnes and Staphylococcus epidermidis contributions to the innate immunity
functions. ................................................................................................................................................ 44
Table 4. Dermatological diseases associated with skin microbiota alterations. ........................... 47
Table 5. TLR-microbiota associations in main cutaneous diseases. ............................................ 135
List of figures
Figure 1. Performing a microbiota study. .. ......................................................................................... 3
Figure 2. 16S rRNA marker gene. ........................................................................................................ 7
Figure 3. 2nd generation sequencing platforms and chemistries. .................................................... 10
Figure 4. 3rd generation sequencing platforms and chemistries.. ................................................... 12
Figure 5. Overview of the bioinformatics workflow used in this thesis to analyze the
microbiota. ............................................................................................................................................. 15
Figure 6. Alpha diversity plots. Red and blue represent two different biological categories. .. . 19
Figure 7. Unweighted and Weighted UniFrac metrics plots.. ........................................................ 20
Figure 8. The PICRUSt workflow... ................................................................................................... 21
Figure 9. Hair coat phenotypes on dogs. .......................................................................................... 24
Figure 10. Skin anatomy and main cells. ............................................................................................ 25
Figure 11. Skin structure: main layers, appendages and associated microbiota.. ......................... 26
Figure 12. Skin in homeostasis and inflammation states................................................................. 28
Figure 13. The hourglass shape of innate immune response.. ....................................................... 29
Figure 14. TLRs and their main ligands. ............................................................................................ 30
Figure 15. Evolution forces acting on TLR genes.. ......................................................................... 32
Figure 16. Human skin microhabitats and their associated microbiota.. ...................................... 36
Figure 17. Skin microbiota composition per skin site.. ................................................................... 37
Figure 18. Temporal stability of the human skin microbiota.. ....................................................... 38
Figure 19. Microbial community of the skin using shotgun metagenomics................................. 39
Figure 20. Skin mycobiota composition on healthy dogs per skin site. ........................................ 40
vii
Figure 21. Canine skin microbiota is potentially shaped by both host and environmental
factors. ................................................................................................................................................... 42
Figure 22. Skin microbiota immune functions on health.. .............................................................. 43
Figure 23. Dynamics of microbial interaction at the skin surface... .............................................. 44
Figure 24. Staphylococcus epidermidis cross-talk with innate immunity.. ............................................ 45
Figure 25. Dermatological diseases associated with dysbiosis on skin microbiota.. ................... 46
Figure 26. Pre- and probiotics for the skin. ...................................................................................... 50
Figure 27. Alternative scenarios of microbial dynamics across different healthy individuals.... 52
Figure 28. Genetic variation on canine Toll-like Receptors.. ....................................................... 134
Figure 29. Example of an interaction between host genome, microbiota and environment:
LCT gene, Bifidobacterium and dairy consumption.. ................................................................... 136
Figure 30. Microbiota of healthy dogs depending on the skin site. ............................................. 138
Figure 31. Skin microbiota profile of healthy dogs at phylum level. . ......................................... 139
Figure 32. Dog skin microbiota on atopic dermatitis. ................................................................... 140
Figure 33. Dogs interacting with the environment. ....................................................................... 143
Figure 34. Experimental approaches to perform a microbiota study. ........................................ 144
Figure 35. Taxonomic coverage at phylum level for bacteria of two universal 16S rRNA gene
primer sets. ........................................................................................................................................... 145
Figure 36. Evolution on the accuracy of Oxford Nanopore Technologies sequencing kits. .. 148
Figure 37. Coverage of the tree of life.. ........................................................................................... 149
Figure 38. Comparison of universal bacterial barcodes: 16S rRNA and Cpn60.. ..................... 150
viii
1
1. Introduction
In the present section we will provide an extensive overview of the topics covered in this
thesis.
The first part will provide the technical framework needed to fully understand a microbiota
study, from the experimental design, to the sample collection and extraction, further
processing and sequencing to the final bioinformatics analyses and interpretation of the
results.
The second part will provide the biological background to understand the object of this study,
which is the skin. It will review skin anatomy and ecology on dogs, as well as genetic variability
in genes responsible for skin innate immunity.
The final part of the introduction will review the present state of the art of skin microbiota on
human and dogs, both in health and in disease.
2
1.1. Conducting a microbiota study
Microbiome studies have boosted since the apparition of next-generation sequencing (NGS)
techniques that allow sequencing DNA massively and in parallel to obtain a huge amount of
data (Kuczynski et al., 2011) (Section 1.1.2). The microbiome is defined as the microbial
community (bacteria, virus, fungi, etc.) that inhabits a specific environment (Box 1) (Marchesi
and Ravel, 2015). Thus, a microbiome sample is synonymous of thousands of different DNA
targets and NGS is the key tool to understand all this complexity. See Box 1 for a collection of
the main key concepts on microbiome research.
Microbiome studies can follow either an amplicon-based approach or a shotgun whole
genome sequencing (WGS) one (Figure 1a). On one hand, amplicon-based approaches aim to
describe the taxonomic composition and diversity of a community amplifying some regions of
16S rRNA marker gene, which is highly conserved among prokaryotes but also has
hypervariable regions that differ among taxa (Section 1.1.1.1.3). On the other hand, WGS
approach aims to sequence all the DNA present within an ecological niche, thus seeing the
entire community (bacteria, archaea, fungi, virus, etc.) with all their genes (Kuczynski et al.
2011; Grice and Segre 2012). This approach gives both taxonomical and functional
information although it is more expensive and computationally challenging (Luo et al., 2013).
Once the design of the study is clear and well defined, samples need to be collected using a
consistent method from the defined skin site/s. Then, researchers should choose the most
adequate DNA extraction protocol and PCR primer sets for their specific environment, if
following an amplicon-based approach. After that, DNA is ready to be sequenced using NGS
techniques. The last step is analyzing the output data using several bioinformatics tools (Figure
1b).
In this thesis we have chosen an amplicon-based approach (16S rRNA gene) to study skin
microbiota in healthy dogs. In the following sections we will provide a more detailed overview
of the workflow to follow when performing an amplicon-based microbiota study of skin
samples. We will review step by step the process, from the experimental considerations and
the main NGS techniques to the data analysis.
3
Figure 1. Performing a microbiota study. In a) main approaches depending on the objective; (from Grice and
Segre 2012). in b) main steps of a skin microbiota study (adapted from Kong et al. 2017).
4
Box 1. Summary of the main nomenclature used on microbiota studies.
Microbiome Collection of microorganisms together with their genes,
in a defined microenvironment
Microbiota Collection of microorganisms in a defined
microenvironment
Metagenome Collection of genes and genomes present in a defined
microenvironment obtained with WGS approaches.
Whole genome shotgun
sequencing (WGS)
From each defined microenvironment DNA is
collected, which can contain bacteria, fungi, viruses and
even host DNA. This DNA is randomly sheared into
small pieces and sequenced. Afterwards, they are
assembled into continuous longer sequences.
16S rRNA sequencing 16S rRNA gene is universal among prokaryotes and can
be used to taxonomically classify bacteria. This
taxonomic resolution comes from its gene structure:
nine hypervariable regions surrounded by highly
conserved regions.
Operational Taxonomic Unit
(OTU)
Cluster of microorganisms that have similar DNA
sequences on a taxonomic marker gene at specific
threshold (97%, 99%, etc.). They are usually
representing a specific bacterial species or taxon.
Taxonomic microbial
composition
Relative abundances of different microorganisms of the
microbiota.
Alpha diversity Measures the number and distribution of OTUs (~taxa)
within a unique sample.
Richness Alpha diversity metrics that consider only the number
of OTUs (e.g. Observed species)
Evenness Alpha diversity metrics that consider the number of
OTUs and their relative abundances (e.g. Shannon
index)
Beta diversity Measures the similarity (~shared OTUs) between
bacterial communities. It computes distance matrices
that can be plotted.
Unweighted UniFrac Beta diversity metrics to assess divergences in
microbiota composition (phylogeny and
presence/absence of OTUs).
Weighted UniFrac Beta diversity metrics to assess divergences in
microbiota structure (phylogeny, presence/absence of
OTUs and relative abundances).
5
1.1.1. Experimental considerations
Microbiome studies using next-generation sequencing need to be designed considering many
technical variables (Kuczynski et al., 2011; Rogers and Bruce, 2010). Researchers have seen
divergences on microbiome composition depending on: sample collection and storage
(Dominianni et al., 2014); DNA extraction protocols (Wagner Mackenzie et al., 2015;
Wesolowska-Andersen et al., 2014); primers chosen for amplification (Kuczynski et al., 2011;
Meisel et al., 2016); sequencing platform used (Castelino et al., 2017; Clooney et al., 2016;
Fouhy et al., 2016) or even clustering method employed on the bioinformatics analyses
(Kopylova et al., 2016; Schmidt et al., 2014, 2015).
Skin microbiota samples are more challenging to process than those from other body sites
(Kong et al., 2017). Thus, researchers have to deal with some extra factors such as: low
microbial biomass linked with a higher risk of sample contamination (Salter et al., 2014); or
the fact that skin has different microenvironments and harbors site-specific microbiota
(Costello et al. 2009; Grice et al. 2009). At least, sample collection methods (Chng et al. 2016;
Grice et al. 2008) or storage conditions (Lauber et al., 2010) seem not to be affecting
significantly skin microbiota structure or composition.
In this dissertation most of the experimental factors were chosen following the recommended
procedures of the Human Microbiome Project consortium (Human Microbiome Project
Consortium., 2012).
1.1.1.1. Microbiota study design
The first step when performing a microbiota study is the study design, which needs to be
appropriate to answer the research question. On one hand, cross-sectional studies including
different skin sites allow describing and assessing skin microbiota variability and diversity on
healthy individuals. Cross-sectional studies can also be case-control studies for assessing skin
microbiota changes during a dermatological disease or any other disruption that affects the
skin. On the other hand, longitudinal studies will allow assessing skin microbiota stability
through time, either under normal conditions or under any disruption (disease, environmental
change).
Cross-sectional studies targeting healthy skin are found in the early stages of the microbiota
research and allow creating a background and determining the normal variability of the skin
microbiota. In human skin microbiota, first studies assessed the normal variability on healthy
individuals using a cross-sectional approach (Grice et al. 2009; Costello et al. 2009). On dog
skin microbiota, only one study aimed to assess the normal variability of different skin sites in
healthy dogs (Rodrigues Hoffmann et al., 2014). Later on, we can find more case-control and
longitudinal studies aimed to assess altered states.
Besides choosing an appropriate study design, one should also define the target population.
To do that, the researcher needs to define which metadata to collect, inclusion and exclusion
criteria, and sample size.
6
Assessment of the metadata is really important on a microbiota study, to account for possible
confounder factors. Commonly collected metadata include age, sex, antibiotic use, and
sampling sites, but also other factors may influence the skin microbiome, such as cohabitation,
hygiene, season, time of day, country of birth, mode of delivery, or diet (Kong et al., 2017).
Moreover, inclusion and exclusion criteria that define the study population should be very well
defined and also the appropriate cohort sample size to reach significant results should be
previously estimated (Kong et al., 2017). Some approaches and tools have been proposed to
assess the suitable sample size (Kelly et al., 2015; La Rosa et al., 2012)
(https://github.com/biocore/Evident).
1.1.1.2. DNA extraction
Microbiota harbors a wide range of microorganisms with different characteristics, so a robust
protocol to extract DNA of all the representatives is needed. For example, gram-positive
bacteria present a cell wall with thick layers of peptidoglycan difficult to break. These
differences in cell wall composition can cause bacterial cell lysis to be less efficient on gram-
positive bacteria, and that in turn can distort the apparent microbiota composition and
diversity (Kong et al., 2017; Yuan et al., 2012).
Thus, to avoid this bias different researchers have evaluated the effect of adding a mechanical
disruption step by bead beating to the conventional DNA extraction procedure, which had led
to improved results (Albertsen et al., 2015; Santiago et al., 2014; Sergeant et al., 2012; Walker
et al., 2015; Yuan et al., 2012). After the lysis, DNA can be purified by alcohol precipitation or
by binding to affinity columns.
1.1.1.3. Amplicon-based approach: 16S RNA marker gene
The 16S small ribosomal subunit gene (16S rRNA) is the most widely used marker gene for
microbiota studies(Kuczynski et al., 2011), due to: (1) it is a gene ubiquitously found in
bacteria and archaea; (2) its structure includes both conserved regions, which can be used for
designing “universal” amplification primers, as well as nine hypervariable regions (V1-V9),
which can be effectively used to distinguish among taxa (Clarridge, 2004) (Figure 2); (3) it has
several large databases of reference sequences and taxonomies, such as Greengenes (DeSantis
et al., 2006; McDonald et al., 2012), SILVA (Quast et al., 2013) or the Ribosomal Database
Project (Cole et al., 2014).
7
Figure 2. 16S rRNA marker gene. a) 16S rRNA gene secondary structure with hypervariable regions indicated (from Yarza et al. 2014); and b) 16S rRNA gene scheme with the most widely used primers for microbiota assessment (modified from Kuczynski et al. 2011).
8
When working with 16S rRNA gene to assign taxonomy, different similarity thresholds are
applied to define a bacterial species. Archaeal and bacterial species were classically defined as a
group of strains with: (a) certain degree of phenotypic consistency, (b) 70% of DNA–DNA
hybridization and (c) at least 97% of gene-sequence identity on their 16S rRNA (Gevers et al.,
2005). Later, another characteristic was added: (d) 94 - 96% of average sequence identities of
shared genes (Richter and Rossello-Mora, 2009).
When focusing on 16S rRNA gene to define a species, the 97% identity cut-off was not
enough within some bacterial genera to discriminate among different species and several
researchers proposed that a 98.7% similarity threshold was more adequate (Stackebrandt and
Ebers, 2006; Yarza et al., 2014). This identity cut-off should probably be adapted within each
bacterial genus (Rossi-Tamisier et al., 2015).
The most commonly used strategy to assess microbiota composition is sequencing specific
hypervariable regions of 16S rRNA gene and clustering the sequences using a similarity
threshold of 97%. Thus, many authors assessed the specificity and universality of several 16S
primers sets (Chakravorty et al., 2007; Jumpstart Consortium Human Microbiome Project
Data Generation Working Group, 2012; Klindworth et al., 2013; Kuczynski et al., 2011;
Mizrahi-Man et al., 2013). These 16S hypervariable regions are sequenced using NGS
platforms, which are classified as 2nd or 3rd generation platforms (reviewed in Section 1.1.2),
depending on the sequencing technology they use.
On one hand, 2nd generation sequencers have sequencing length limitations, so researchers
have to choose primers targeting specific hypervariable regions to obtain short amplicons.
Among the different options, the most universal primer set is F515-R806 (region V4) that
captures both bacteria and archaea. However, it fails to amplify Propionibacterium, which is an
abundant genus on the skin, thus in this dissertation we have worked with F27-F338 (region
V1-V2), which was more suitable for our skin microbiota samples (Kuczynski et al., 2011;
Walters et al., 2011).
On the other hand, 3rd generation sequencers can sequence long reads, so researchers choose
amplifying the full (or almost full) length 16S rRNA gene using several universal primers
(Klindworth et al., 2013). This approach gives a better and more accurate assessment of
taxonomic diversity (Yarza et al., 2014). Here in this dissertation we used mainly primer set
F27-R1492 (region V1-V9) and also F27-R1391 (region V1-V8) when testing Oxford
Nanopore MinION™ 3rd generation sequencer. These two primer sets have low non-coverage
rates even at phylum level being a good choice to assess microbiota diversity (Mao et al.,
2012).
9
1.1.2. Next generation sequencing (NGS)
Once the sample has been successfully processed, the next step is transforming the DNA into
data. The development of NGS has allowed obtaining huge amount of data in short periods
of time, although with higher associated error rates (~0.1–15%) when compared to classical
Sanger sequencing (Goodwin et al., 2016).
Main aspects to consider when choosing the appropriate sequencing platform for a specific
study are: length of the reads; number of the reads (sequencing depth); and error rate and type
(Vincent et al., 2016). The length of the reads is linked to the amount of information obtained
from a single molecule, which is particularly significant when the aim is to phylogenetically
differentiate 16S rRNA gene fragments (Jumpstart Consortium Human Microbiome Project
Data Generation Working Group, 2012; Schloss et al., 2016a). The total number of reads is
the most important parameter for quantitative applications (Vincent et al., 2016), such as
determining the microbiota diversity and composition. The combination of read length and
number of reads defines the throughput of an instrument in number of bases per run. Both
the error rate and the different types of errors are specific of each sequencing technology and
can be minimized with an increased coverage (Vincent et al., 2016)(Goodwin et al., 2016).
Conventionally NGS platforms have been divided in two groups: 2nd and 3rd generation
platforms. 2nd generation sequencing aim is to obtain large amounts of data by massive
sequencing in parallel millions of short reads at an affordable cost. The sequencing process
requires a previous PCR-amplification of the DNA. In contrast, 3rd generation platforms use a
single-molecule real-time sequencing approach, which means that they can sequence individual
DNA molecules avoiding the necessity of PCR amplification and its associated bias. Another
advantage is their ability to sequence long fragments (Glenn, 2011; Goodwin et al., 2016;
Mardis, 2017).
Some of the most commonly used platforms of 2nd generation sequencing are 454, Illumina
and Ion Torrent, which have some common and distinctive traits (Figure 3). The first step in
the sequencing process is the library preparation, were the DNA is fragmented and the
sequencing adaptors ligated. The following step is the clonal amplification of the individual
DNA fragments that can be bead-based (454 and Ion Torrent) or solid-state (Illumina). The
last step is the sequencing process itself.
Both 454 and Ion Torrent perform a bead-based amplification prior to sequencing, where
each DNA molecule is immobilized in a single bead and clonally amplified using emulsion
PCR (emPCR) (Figure 3b.1) (Dressman et al., 2003). In the case of Illumina a solid-state
amplification is used with covalently bound forward and reverse primers where DNA
molecules attach and clonally amplify forming clusters of sequences (Figure 3c.1).
10
Figure 3. 2nd generation sequencing platforms and chemistries. a) Chronological timeline with the launching dates of the main platforms. In b) 454 and Ion Torrent sequencing technologies: b.1) bead-based clonal amplification, prior to b.2) 454 sequencing or b.3) Ion Torrent sequencing. In c) Illumina sequencing technology: c.1) solid-state clonal amplification (bridge amplification) prior to c.2) sequencing with Illumina and its sequencing strategy (Figure modified from Goodwin, McPherson, and McCombie 2016)
11
The three 2nd generation platforms reviewed here use sequencing by synthesis (SBS)
approaches, so they rely on DNA polymerases. They differ mainly on the sequencing
chemistry and the produced secondary signals.
- 454 pyrosequencing (Figure 3b.2) was the first 2nd generation platform that
successfully achieved commercial introduction, and it was released on 2004 (Mardis,
2017). Once a polymerase incorporates a dNTP into a strand, it releases a
pyrophosphate molecule that produces a bioluminescence signal through an enzymatic
cascade. Differences on light signal intensity will indicate how many dNTPs are
incorporated, which sometimes can lead to homopolymer errors (Goodwin et al.,
2016).
- IonTorrent PGM (Figure 3b.3) was the first 2nd generation platform to sequence
without optical sensing (Rothberg et al., 2011). Once a polymerase incorporates a
nucleotide to a sequence, a proton (H+) is liberated producing a measurable change in
the pH. The pH change is detected through a sensor and it is proportional to the
number of nucleotides incorporated. However, it is not perfectly linear and presents
certain problems with homopolymers (Goodwin et al., 2016; Morey et al., 2013).
- In Illumina platforms (Figure 3c.2), each nucleotide is labeled with a base-specific
cleavable fluorophore and then blocked, which allows incorporating one nucleotide
per cycle. The posterior identification is achieved through total internal reflection
fluorescence (TIRF) microscopy (Goodwin et al., 2016). Among the different Illumina
platforms, the most commonly used in microbiota studies is MiSeq.
Main limitations of 2nd generation platforms in a microbiota project are: i) the short read
length, which makes difficult the correct taxonomic classification up to species level using
specific hypervariable regions of 16S rRNA (Schloss et al., 2016a); and ii) the need of a PCR-
amplification step previous to sequencing, which can bias the original microbiota composition
of an specific environment. 3rd generation sequencers emerged to overcome some of these
limitations. Nowadays, the main technologies and sequencers in the market are: Single
Molecule Real Time Sequencing from Pacific Biosciences (PacBio RS II platform) and Single-
Molecule Nanopore sequencing from Oxford Nanopore Technologies (ONT: MinION,
GridION and PromethION platforms) (Glenn, 2011; Goodwin et al., 2016; Heather and
Chain, 2016; Mardis, 2017). The main aim of these platforms is sequencing single molecules at
real-time, without short-length limitation and high-throughput, although lower than 2nd
generation platforms.
12
Figure 4. 3rd generation sequencing platforms and chemistries. a) Chronological timeline with the launching of the main platforms. In b) Pacific Biosciences and c) Oxford Nanopore Technologies sequencing. (Figure modified from Goodwin, McPherson, and McCombie 2016)
13
Sequencing technologies behind these two sequencers are very different: while PacBio
continues using sequencing by synthesis, Oxford Nanopore Technology relies on direct-
sequencing through nanopores. More detailed:
- Pacific Biosciences (PacBio, http://www.pacb.com/) platform is based on the single-
molecule real-time (SMRT) sequencing approach (Eid et al., 2009). DNA molecules
will be sequenced in SMRT-bell structure, which consists on a double-stranded region
(DNA insert of interest) ligated with a single-stranded hairpin loop on both ends that
provides a binding site for the primer. This structure enables multiple passes of the
polymerase per DNA molecule, increasing the accuracy by creating a circular
consensus sequence (CCS) (Travers et al., 2010). Sequencing reactions take place in
picoliter wells called zero-mode waveguides (ZMW) (Levene et al., 2003) that have in
their transparent bottom a fixed polymerase. By having a constant location of
nucleotide incorporation, the optical system can focus on a single molecule and once a
fluorophore is cleaved it diffuses away from the sensor. Thus, each dNTP
incorporated on a single-molecule template emits a light that is continuously recorded
and later basecalled (converted) to a DNA read (Goodwin et al., 2016).
- Oxford Nanopore Technologies (ONT, https://nanoporetech.com) launched in 2014
a small, USB-like, portable sequencer based on nanopore sequencing: MinION™ (Ip
et al., 2015). Unlike other technologies, nanopore sequencing directly detects the DNA
composition of a native single strand DNA molecule rather than a secondary signal
(light, pH, colour, etc.) derived from the synthesis of a complementary strand
(Goodwin et al., 2016). A DNA strand is sequenced as it passes through the nanopore
while electrical current passes through the pore (Clarke et al., 2009). DNA molecules
passing through the nanopore produce blockages in electrical conductivity that can be
used to discriminate individual nucleotides (Olasagasti et al., 2010; Wang et al., 2015).
Each subset of nucleotide bases (k-mer) translocating through the nanopore produces
a unique shift of voltage called squiggle, which is the raw signal that will be basecalled to
a DNA read. DNA template used in ONT platforms presents a leader sequence that
interacts with the nanopore to direct the DNA through it, and a hairpin that joins the
template and complement DNA strands when performing bidirectional sequencing
with 2D chemistry (Goodwin et al., 2016)(Goodwin et al., 2016). Since May 2017, 2D
sequencing chemistry has been discontinued and 1D2 is replacing it. This new
sequencing strategy does not require a hairpin anymore, avoiding secondary structure
problems and improving accuracy and throughput (Brown, 2017).
14
Several reviews aimed to compare different sequencing technologies available on the market
from different perspectives (van Dijk et al., 2014; Glenn, 2011; Goodwin et al., 2016; Loman
et al., 2012; Morey et al., 2013; Reuter et al., 2015; Vincent et al., 2016). In general, 3rd
generation sequencers allow obtaining longer reads but they present higher error rates when
compared to 2nd generation sequencers. Some other divergences can be found in Table 1.
Table 1. Main sequencing platforms used in microbiota studies and their main characteristics.
Platform Type of instrument
Amplification Sequencing chemistry
Secondary signal
Read length (bp)
Ion Torrent PGM Benchtop emPCR Synthesis H+ (pH change) 200-400
Illumina MiSeq Benchtop solid-state Synthesis Light 150-300
454 Roche Benchtop emPCR Synthesis Light 400-650
PacBio RS High-end machine - Synthesis Light up to 60,000
MinION Portable - - - ultra-long reads*
*Nanopore sequencing read length is only limited by high-molecular weight DNA extraction, the longest read
obtained at date 16/05/2017 is close to 1 Mb by Josh Quick & Nick Loman (Quick and Loman, 2017).
In this dissertation, we work with data from Ion Torrent PGM™ to assess skin microbiota
composition (Chapter 3.2 and 3.3). Moreover, we assess the potential of nanopore sequencing
at technical level for microbiota studies using MinION™ by ONT (chapter 3.4).
1.1.3. Bioinformatics analysis
The starting point of a microbiota bioinformatics analysis is the raw data obtained from the
sequencing platform. In this dissertation, we have mainly worked with output from Ion
Torrent PGM™ (as a 2nd generation sequencer). Moreover, we have also tested nanopore
sequencing using MinION™ (as a 3rd generation sequencer). For more details on nanopore
data processing see Chapter 3.4.
In this section, we will cover the main analytical steps performed in a microbiota study
illustrating them with the bioinformatics tools and the metrics used in this thesis.
Usually microbiota studies rely on pipelines that gather together many different bioinformatics
tools. In this thesis, we mainly used Quantitative Insights Into Microbial Ecology (QIIME
v1.9.1) (Caporaso et al., 2010a) and in Chapter 3.3 also VSEARCH in some steps (Rognes et
al., 2016a). Another equivalent and widely used pipeline is mothur although we have not used it
(Schloss et al., 2009).
The analytical workflow is divided in three main steps: (i) pre-processing of the sequences, (ii)
operational taxonomic unit (OTU) picking and OTU table building and (iii) downstream
analysis (Figure 5).
15
This taxonomic-based analysis can be completed with tools predicting the functional potential
of the microbial community. We have used Phylogenetic Investigation of Communities by
Reconstruction of Unobserved States (PICRUSt) (Langille et al., 2013), but other tools such as
Tax4fun (Aßhauer et al., 2015) and PanFP (Jun et al., 2015) can be used for the same aim.
Finally, Linear Discriminant Analysis Effect Size (LEfSe) (Segata et al., 2011)) software is
usually used to assess statistically which taxa is under or overrepresented between different
groups of samples.
Figure 5. Overview of the bioinformatics workflow used in this thesis to analyze the microbiota.
16
1.1.3.1. Pre-processing step
Each Ion Torrent PGM™ run contains a pool of several barcoded samples, with an average
length of 350 bp for primers 27F-R338R targeting V1-V2 regions of 16S rRNA gene. Raw
data from Ion Torrent PGM™ is given demultiplexed, which means separated by barcodes,
thus we have a unique file per sample included in the run.
The pre-processing steps are performed using Quantitative Insights Into Microbial Ecology
(QIIME v1.9.1) software (Caporaso et al., 2010a). We need two input files, one containing the
DNA sequences and the other its associated metadata. Whereas the DNA sequences are the
output of Ion Torrent, the mapping file with all the metadata needs to be created by the
researcher. This should contain all the technical information and metadata about the samples:
sample ID, forward and reverse primers, treatment, description, age, or any other relevant
information.
The following step is merging the sequencing file with the mapping file for each sample.
Reading the information on the mapping file, the script trims the primers and labels the
sequences with their sample ID.
Moreover, this script performs the initial quality control of the sequences discarding those
ones shorter than 300 bp, with quality phred scores below 25, with mismatches on the primer
and other default parameters (script: split_libraries.py).
1.1.3.2. Operational taxonomic unit (OTU) picking and OTU table building
The basic unit when working with 16S data is the Operational Taxonomic Unit (OTU), which
is constituted by a group of sequences that share a certain percentage of similarity. Usually
sequences with 97% of similarity are clustered together constituting an OTU, but other
percentages have been proposed.
There are two main strategies to pick OTUs: de novo, which clusters the sequences among
them considering a specific percentage of similarity (97% of similarity in our studies); and
closed, which clusters the sequences against a reference database (previously clustered at the
similarity threshold chosen). The first approach allows all the sequences to be clustered,
although it is usually computationally expensive. The second approach is faster, but sequences
that are not in the database are excluded for further analysis. QIIME preferred option is a
combined one, which is called open reference: It first clusters the sequences against a database
and the ones that do not match are clustered de novo among them (Rideout et al., 2014). We
have used both open and closed reference approaches in Chapter 3.2, see further details in
material and methods section (script: pick_open_references_otus.py). When working with
VSEARCH (Rognes et al., 2016b) we have used the de novo strategy because it runs faster than
its homologous in QIIME. See further details in Chapter 3.3, in the material and methods
section.
Chimeras are amplification artifacts really common in microbiota surveys (Ashelford et al.,
2005), due to the large amount of PCR targets within a single reaction. Thus, an aborted
17
extension can work as a primer for another target forming a chimeric artifact. If chimeras are
not removed from the data, they are incorrectly identified as novel taxa and that leads to an
inflated microbial diversity (Ashelford et al., 2005; Haas et al., 2011). Chimeric artifacts can be
minimized by the correct choice of 16S rRNA primers and an adjusted number of PCR cycles
among others (Schloss et al., 2011). Moreover, once you have already sequenced the data,
several bioinformatics tools have been designed to identify chimeric sequences and remove
them. Here we have used ChimeraSlayer (Haas et al., 2011) after OTU picking on QIIME and
UCHIME (Edgar et al., 2011) in VSEARCH pipeline.
Once we have picked OTUs and removed the invalid and low-quality sequences from the
analysis, we assign the taxonomy to our sequences as well as align them to create a
phylogenetic tree. Taxonomic assignment of the OTUs is performed using the RDP Classifier
(Wang et al., 2007) against Greengenes v13.8 database (DeSantis et al., 2006; McDonald et al.,
2012). Alignment of sequences is performed using PyNast (Caporaso et al., 2010b) as default
in QIIME pipeline.
1.1.3.3. Downstream analysis
1.1.3.3.1. Taxa summary
The simplest way to describe a microbial community is with a list of the bacteria together with
their abundances, which are usually plotted in bar graphs.
V1 and V2 hypervariable regions of 16S allow classification of the sequences mostly up to
family level, some of them even up to the genus level. Taxa analysis can be summarized at a
specific taxonomic level or can be global.
1.1.3.3.2. Microbial ecology measures: alpha and beta diversity
Microbial diversity is assessed using metrics that can take into account: OTU counts, relative
abundances and/or phylogenetic information. When bacterial community diversity is assessed
within a community, it is called alpha diversity. On the other hand, when bacterial diversity is
compared and assessed among different communities, it is called beta diversity.
Many different ecological indexes have been described for assessing both alpha and beta
diversity, which can be classified depending on which properties they take into account: (1)
qualitative, when only presence/absence data is considered; or (2) quantitative, when also
relative abundances are taken into account; they can also be (3) species-based, when different
taxa are treated as equally related; or (4) divergence-based, when phylogenetic distances among
each pair of taxa are considered (Lozupone and Knight, 2008, 2007).
In this dissertation we choose two species-based methods to assess alpha diversity: Observed
species (qualitative) and Shannon index (quantitative) (Shannon and Weaver, 1949). On the
other hand, for beta diversity analysis we choose two divergence-based methods: Unweighted
18
UniFrac (qualitative) (Lozupone and Knight, 2005) and Weighted UniFrac (quantitative)
(Lozupone et al., 2011).
Alpha diversity
Alpha diversity assesses diversity within a community. As a qualitative index we used
Observed Species metric, which assesses richness of the sample by simply counting the unique
OTUs found. As a quantitative index we used Shannon index (Shannon and Weaver, 1949),
which assesses the evenness of a sample. Shannon index value gets larger as the number of
species increases and as the distribution of species becomes even.
Alpha diversity is usually highly dependent on sequencing depth. For that reason, it is
common to use rarefaction curves to show the cumulative number of species as a function of
sampling depth (Figure 6a and 6b)(Hughes and Hellmann, 2005). Consequently alpha diversity
should always be given at a specific sequencing depth. Alpha diversity values can be
represented in tables, or in plots similar to those shown in Figure 6.
Finally, as an example to better understand the concepts, we should now have a look at Figure
6. We have a study cohort of nine samples that belong to two groups (A in red and B in blue).
In Figure 6a we can see the rarefaction curves of all the samples included in one study, and
despite presenting uneven sequencing depths, all reach the sequencing depth of 8,300 reads
per sample. So, we decide to work at this sequencing depth and now we group the samples
from the same group together and create a second rarefaction plot (Figure 6b). Finally, we can
visually compare groups using boxplots (Figure 6c) and even assess statistical significance
using non-parametric tests, e.g. Monte Carlo permutation test.
19
Figure 6. Alpha diversity plots. Red and blue represent two different biological categories. a) Alpha diversity rarefaction plot of all the samples at their own sequencing depth; all the samples reach 8,300 sequences per sample so further analyses are performed at this depth. b) Alpha diversity rarefaction plot of the two groups. c) Boxplots representing alpha diversity values distribution within each group.
Beta diversity
Beta diversity assesses differences among bacterial communities computing distance matrices
that can be combined with multivariate statistical techniques, such as principal coordinate
analysis (PCoA) and hierarchical clustering. These statistical analyses allow plotting the results
graphically as well as detecting patterns and clusters of samples.
Because sample depth can affect beta diversity results and to prove its robustness, it is always
recommended to apply a jackknifing protocol. Jackknifing consists on subsampling randomly
and repeatedly the initial sample using a specific sequencing depth (Zahl, 1977).
20
In this dissertation, we choose UniFrac metrics that takes into account the phylogenetic
distance among bacteria. Thus, two communities sharing similar phylogenetic lineages will be
more similar, despite not sharing the same exact species.
UniFrac distance matrix (Lozupone and Knight, 2005) measures the phylogenetic distance
between two collections of sequences as the fraction of branch length in a phylogenetic tree
that leads to descendants of one sample or the other, but not both. Besides phylogenetic
information, when these metrics consider only the presence and absence of the different
bacteria found in the community it is called Unweighted UniFrac, whereas when they take into
account relative abundances it is called Weighted UniFrac (Figure 7) (Lozupone et al., 2007).
Figure 7. Unweighted and Weighted UniFrac metrics plots. Beta diversity represents distances among the bacterial communities using either UPGMA trees or PCoA plots. Bacterial community B has the same bacterial species as bacterial community C, A presents one more species. On the other hand A and B present similar relative abundances of the main bacteria, whereas C presents a predominant bacterium. a) Unweighted UniFrac, which only considers composition and phylogeny, B and C will be more similar and b) Weighted UniFrac, which also considers relative abundances, A and B will be more similar.
21
To assess the extent and significance of the clustering of the samples regarding a specific
grouping variable, non-parametric statistical tests such as ANOSIM (analysis of
similarities)(Clarke, 1993) and adonis are commonly used. ANOSIM test give an R value that
when it is closer to +1 means that the dissimilarity among groups is high, whereas when it is
closer to 0 means that there is no grouping based in that variable. Adonis test computes an R2
value (effect size) that shows the percentage of variation explained by the grouping category.
Both tests also output a p-value to determine the statistical significance.
1.1.3.3.3. Functional prediction using PICRUSt
Profiling 16S rRNA gene is a widely used method to provide insights into microbial
community, although a more complete view would include information about community’s
functional capabilities. This can be achieved with metagenomics approaches, but when only
16S data is available it is also possible by predictive tools such as PICRUSt (Langille et al.,
2013).
PICRUSt needs an OTU table obtained through a closed reference approach, since the
prediction is based on a reference database. The first step is normalizing the OTU table by
dividing the OTU counts by predicted marker gene copy number. The second step is inferring
metagenomes, multiplying the inferred number of OTUs per sample by the predicted gene
content. The final output is a matrix of the gene count with samples as columns and
functional pathways as rows (Figure 8).
To assess the reliability of the software, Langille and colleagues used this tool with the Human
Microbiome Project dataset (Human Microbiome Project Consortium, 2012) obtaining
sufficiently accurate results, even for skin samples (Langille et al., 2013).
Figure 8. The PICRUSt workflow. The metagenome inference workflow takes a closed-reference OTU table, as well as the copy number of the marker gene in each OTU and the gene content of each OTU and outputs a metagenome table (i.e., counts of functional pathways per sample) (Figure adapted from Langille et al. 2013).
22
1.1.3.3.4. Detecting differentially distributed features using LEfSe
In both taxonomic and predicted functions analyses, we usually want to assess which features
are statistically different among groups. To assess that, here in this dissertation we choose to
use Linear Discriminant Analysis (LDA) Effect Size (LEfSe) algorithm (Segata et al., 2011).
LEfSe algorithm first uses the non-parametric factorial Kruskal-Wallis (KW) sum-rank test
(Kruskal and Wallis, 1952) to identify features that are statistically different among biological
classes. Biological significance is subsequently assessed using a set of pairwise tests among
subclasses (when provided). As a last step, LEfSe uses Linear Discriminant Analysis (Fisher,
1936) to estimate the effect size of each differentially abundant feature.
1.1.3.3.5. Building ecological networks using CoNet
The microbiome is a complex ecosystem where microbes compete and cooperate. These
microbial interactions can support health or promote disease. Thus identifying and
characterizing them will allow us to better understand microbial dynamics.
We used CoNet software (Faust and Raes, 2016) in Chapter 3.3, which is implemented as an
application in Cytoscape (Shannon et al., 2003), to obtain an overview of the microbial
dynamics underlying canine skin microbiota.
The input file needed is an OTU table that will be first pre-processed normalizing the data and
removing the taxa with too many zero values, which can lead to spurious interactions. In the
next step, you can select the methods used to compute the networks that can be correlations,
similarities and/or dissimilarities metrics. Most of these association measures allow assigning a
positive or negative sign to a predicted relationship, which will be plotted on the network
using a green or red edge respectively. The recommended approach is to choose multiple
metrics and then combine networks, and if in any edge measures disagree on the sign, it is
discarded (Faust and Raes, 2016).
23
1.2. Overview of the skin
Skin is an anatomical barrier that separates the animal from the outer environment. However,
it is not only an anatomical barrier but also a living barrier, covered by thousands of
microorganisms that cross-talk with the host cells as well as the immune system, maintaining
the homeostasis and equilibrium (Naik et al., 2012a). Besides being a physical and
immunological barrier, skin also percepts different kind of stimuli; facilitates motion and gives
shape; produces adnexa (such as sweat and sebaceous glands, claws or hairs); regulates body
temperature; stores fat, proteins, vitamins or water among others; may indicate the general
health status or sexual identity; gives pigmentation; has antimicrobial properties; produces
vitamin D; and secretes and excretes (Miller et al., 2013).
The object of study in this dissertation is the skin, so we will delve into its anatomy and
physiology as well as into its cellular composition and histology. Finally we will provide some
background on skin innate immune functions and specifically Toll-like Receptors as the first
sensors of microbes.
24
1.2.1. Anatomy and physiology of the canine skin
On dogs, a dense hair coat, also named fur, covers the skin. The hair coat insulates the skin
thermally and sense stimuli, as well as protects the skin against chemical, physical and
microbial damage (Miller et al., 2013). Breeds have been classified regarding their hair coat
length in four main groups: short, wire, long, and curly. These phenotypes and some
combinations and variations have been associated to alleles located in three genes: FGF5,
RSPO2 and KRT71 (Figure 9) (Cadieu et al., 2009).
Figure 9. Hair coat phenotypes on dogs. Different combinations of alleles located on FGF5, RSPO2 and KRT71 genes produce most of the hair coat phenotypes on dogs (Figure from Cadieu et al. 2009)
Both the skin and the hair coat vary dynamically within individual canine species in quantity
and quality depending on many factors such as the age, the sex, the skin site, the breed and
even the individual (Miller et al., 2013). Daily growth rate of the hair shaft is both season
dependent –being greater during the colder season– and site dependent (Al-Bagdadi, 2013).
In general, skin thickness decreases dorsally to ventrally on the trunk and proximally to distally
on the limbs. Thus, the hair coat is usually thickest over the dorsolateral regions of the body
and thinnest ventrally, on the concave part of the pinnae and on the undersurface of the tail
(Miller et al., 2013; Scott et al., 1995).
The skin of haired mammals is usually acidic, however normal pH values on dog skin have
been reported to range from 4.84 to 9.95 (Miller et al., 2013). A dynamic study assessing pH
on healthy dogs (Ruedisueli et al., 1998) reported that: 1) pH values differ day to day and
among skin sites; 2) males and spayed females had significantly higher pH values on all skin
sites than females and intact females, respectively; 3) black Labrador retrievers had
significantly higher pH values than yellow ones; and 4) some breeds had significantly different
pH values when compared to other breeds.
25
1.2.2. Histology of the canine skin
Three main cellular layers constitute the dog skin: the epidermis, the dermis and the
hypodermis (or subcutis). Moreover, the skin presents invaginations and appendages that go
from the deep dermis to the skin surface, being the main ones sweat and sebaceous glands and
hair follicles (Miller et al., 2013).
1.2.2.1. The epidermis
The epidermis is mostly constituted by keratinocytes (~85% of the epidermal cells), followed
by Langerhans cells (~3-8%), melanocytes (~5%) and Merkel cells (~2%) (Figure 10) (Miller
et al., 2013). Keratinocytes and Langerhans cells are members of the skin immune system,
being the keratinocytes the first line of defense against invading pathogens (Suter et al., 2009)
whereas Langerhans cells are antigen-presenting cells (White and Yager, 1995). Melanocytes
are the pigment (melanin) producers and Merkel cells are mechanoreceptors involved in light
touch sensation (Miller et al., 2013).
The epidermis of the dog is usually subdivided into four to five cell layers anchored to the
dermis: the stratum basale, the stratum spinosum, the stratum granulosum, the stratum lucidum and the
stratum corneum (Figure 10). Each epidermic layer is mainly representing different maturation
stages of the keratinocytes (Miller et al., 2013).
Figure 10. Skin anatomy and main cells. The epidermis is constituted by keratinocytes at different stages and they form layers called the stratum basale, the stratum spinosum, the stratum granulosum and the outermost layer, the stratum corneum. Specialized cells in the epidermis include melanocytes, which produce pigment (melanin), and Langerhans cells (Figure modified from Nestle et al. 2009).
Stratum basale is made of a single layer of columnar cells, mostly of them dividing and non-
dividing keratinocytes. As basal keratinocytes reproduce and mature, they move towards the
26
outer layer of skin, initially forming the stratum spinosum. In this layer, keratinocytes are
connected one to other through intercellular bridges. Continuing their transition to the surface
the keratinocytes of the stratum granulosum flatten, their nuclei are reduced and their cytoplasm
appears granular (Miller et al., 2013). Stratum lucidum is only present in footpads and nasal
planum, and it is a fully keratinized compact layer of dead cells (Schwarz et al., 1979). Finally,
the stratum corneum is made of non-viable terminally differentiated keratinocytes known as
corneocytes. These cells (bricks) are embedded in a lipid-rich matrix (mortar) which helps to
hold them together (Leigh and Watt, 1994), as well as retain the water.
1.2.2.2. Appendages and invaginations of the dog skin
In dog skin, the main cutaneous invaginations and appendages are sebaceous glands and sweat
glands (apocrine and eccrine), and the hair follicle. These different cutaneous structures with
their secretions allow creating different microenvironments, which likely influence the
microbiota they harbor as suggested in human skin (Figure 11) (Grice and Segre 2011).
Figure 11. Skin structure: main layers, appendages and associated microbiota. (Figure from Grice and
Segre 2011).
Sebaceous glands are simple or branched alveolar glands distributed throughout all haired skin
in mammals with association to hair follicles. However they are absent in the dog’s footpads
and nasal planum. They are larger and more abundant near mucocutaneous junctions, in the
interdigital spaces, on the dorsal part of the neck and rump, on the chin and on the dorsal tail
(Miller et al., 2013). They secrete sebum that is an oily substance composed mainly by
triglycerides and wax esters (Clarys and Barel, 1995), which bacteria metabolize to convert
them to free fatty acids. The sebum provides a hydrophobic coating of the hair and skin that
27
protects from overwetting and may have a role on heat insulation (Smith and Thiboutot,
2008).
Sweat glands can be epitrichal (apocrine) or atrichial (eccrine) and, unlike humans, they do not play a
significant role in thermoregulation (Cotton et al., 1975). Epitrichal sweat glands are distributed
throughout all haired skin in exception of footpads or nasal planum. Epitrichal sweat has
probably pheromonal and antimicrobial properties (Miller et al., 2013). Atrichial sweat glands
release a watery secretion and are only found in the nasal planum and footpads placed deeply
in the dermis and subcutis (Scott et al., 1995).
The hair follicles of the adult dog are complex and consist of bundles of hairs sharing a
common opening on the skin surface. Usually a primary larger hair is surrounded by few
secondary and thinner hairs forming a hair follicle complex. Each hair has its own sebaceous
gland, but only the primary hairs are associated with an epitrichal sweat gland and an arrector
pili muscle. The contraction of this muscle empties the contents of the glands on the skin
surface (Al-Bagdadi, 2013). There are breed differences regarding both the number of hairs
using a same hair follicle orifice and the density of hair follicles in skin surface.
1.2.3. Skin immunity
In skin we can find complex immunological processes of both innate and adaptive immune
system (Bangert et al., 2011; Pasparakis et al., 2014). Keratinocytes are the first active
participant in the skin immune response, but also Langerhans cells (Nestle et al., 2009) and
even mast cells seem to have a role in skin immunity (Kumar and Sharma, 2010).
The main functions of the innate immune system are detecting pathogens and presenting their
antigens to the adaptive immune system, as well as differentiating self vs non-self cells and
microorganisms. All microorganisms –commensals and pathogens– have structures with
conserved molecular patterns called microbial-associated molecular patterns (MAMPs), when
these ones are specific from a pathogen they are called pathogen-associated molecular patterns
(PAMPs). PAMPs are present in bacteria, fungi, viruses and other parasites and can be nucleic
acids or structural components of the membrane or the envelope. Sensors of pathogens of the
innate immune system are called PAMPs Recognition receptors (PRRs) and major families
include Toll-like Receptors (TLRs), RIG-I like Receptors (RLRs) and Nod-like Receptors
(NLRs) (Kawai and Akira, 2009).
The innate immunity of a healthy individual is capable to defend against pathogens as well as
not reacting against its own cells. However, immunological dysregulations occur and create
inflammation. On one hand, innate immune pathways can be improperly activated by host
cells or endogenous DNA, causing an autoimmune response (Beutler, 2009; Fischer and
Ehlers, 2008). On the other hand, a defect on innate immunity can make someone more
susceptible to infections or even can transform a normal infection to a chronic one (Kumar et
al., 2013) (Figure 12).
28
Figure 12. Skin in homeostasis and inflammation states. a) In a healthy individual, all the microorganisms are in equilibrium and pathogens are controlled with the innate immune system. b) When immunity defects exist or occur, innate immune system can detect own cells and microbiota as pathogens and overreact against them producing an autoimmunity reaction. c) When immunity defects exist or occur, pathogens can easily invade and colonize the skin, innate immune system fails to control the overgrowth and that produces an infection.
This dissertation is focused on microbiota analyses, not innate immunity; however they are
two sides of the same coin. Therefore, we also describe innate immunity of dog skin by
screening polymorphisms on Toll-like Receptor genes and here we present a little overview of
these innate immune receptors.
1.2.3.1. Toll-like Receptors (TLRs)
Toll-like receptors are considered to be the first sensors of microbes, through their microbial
associated molecular patterns (MAMPs) (Werling and Jungi, 2003). The pool of microbes they
have to sense is huge and comes both from the environment and from the own microbiota.
Despite the high variable microbial world, some molecular structures (MAMPs and PAMPs)
are highly conserved across them and are sensed through TLRs (Beutler, 2004; Kumar et al.,
2009) (Figure 13).
TLRs are transmembrane proteins constituted by leucine-rich repeat (LRR) domains, a unique
intramembrane domain and a Toll/Interleukin-1 receptor (TIR) domain. MAMPs and PAMPs
are sensed through the LRR domain, and signals are transduced through the TIR domain,
which is always located in the cytoplasm, in order to activate innate immunity response.
Several crystallography studies elucidated TLRs structure: TLRs act in dimers and create
structures together with their ligands (Gay and Gangloff, 2007; Kang and Lee, 2011; O’Neill
et al., 2013).
29
Figure 13. The hourglass shape of innate immune response. Ten Toll-like receptors (TLRs), four TIR adaptors and two protein kinases are required for most microbial perception (Figure excerpted from Beutler 2004).
Up to date, 10 functional TLRs have been described in humans and 12 in mice (Kawai and
Akira, 2009). Public repositories have sequences for 10 TLRs in dogs.
TLRs have been classically classified depending on their cellular localization and their ligands:
extracellular or intracellular TLRs. TLR1, 2, 4, 5, 6 and 10 are located in the cellular membrane
and usually recognize components of the microorganism cell wall or membrane, whereas TLR
3, 7, 8 and 9 are usually located in intracellular vesicles and recognize mainly nucleic acids
(Kawai and Akira, 2010). Once TLRs detect PAMPs, they send a signal trough an adapter
molecule (MyD88 or TRIF) to initiate in the nucleus the production of transcription factors
(NFκB, IRF3 or IRF7) that will activate pro-inflammatory cytokines and type 1 interferons
(Figure 14) (Kumar et al., 2011; Netea et al., 2012) .
30
Figure 14. TLRs and their main ligands. Extracellular TLRs (above) and intracellular TLRs (below) act in
dimeric form to detect their ligands and initiate the innate immune response that leads to the stimulation of
proinflammatory cytokines.
1.2.3.1.1. Evolution and variability of TLR genes
TLRs are innate immune genes that are highly conserved among different vertebrates (Roach
et al., 2005)(Song et al., 2012). As all the other PRRs, TLR genes evolve under both non-
adaptive and adaptive forces. The non-adaptive evolution is common to all genes within a
species and is produced through genetic drifts, bottlenecks and migration routes. The adaptive
evolution is stronger on PRRs genes and is mainly caused by infectious pressures (Netea et al.,
2012). In general, when novel DNA variants appear in a population they can be under three
types of selection: purifying, positive or balancing (Harris and Meyer, 2006; Quintana-Murci,
2016) (Box 2).
When looking at TLR evolution throughout different vertebrates, it was seen that purifying
selection is mostly acting in intracellular domains that have to give the signal that triggers
immunity; and positive selection is mostly acting in extracellular domains that have to
recognize different pathogens and adapt to different infectious pressures (Werling et al., 2009;
31
Zhou et al., 2007). Despite being highly conserved, signatures of different types of selection
have been detected on TLR genes when comparing different species or even populations that
show the infectious pressures they have undergone (Barreiro et al., 2009; Netea et al., 2012;
Quach et al., 2013).
On human TLRs, different selection forces were described to drive genetic variation. Purifying
selection has been detected in intracellular TLRs (TLR3, 7, 8 and 9) and not in extracellular
TLRs (TLR1, 2, 4, 5, 6 and 10) suggesting that intracellular TLRs have an essential non-
redundant role in host survival (Barreiro et al., 2009; Wlasiuk and Nachman, 2010). On the
other hand, extracellular TLRs evolved under less evolutionary pressure and tolerate more
damaging mutations, which suggests their likely immunological redundancy (Barreiro et al.,
2009). In contrast, Mukherjee and colleagues found that also TLR2 and 4 were under
purifying selection (Mukherjee et al., 2009, 2014). Moreover, another study also reported that
human TLR genes were unequally polymorphic: TLR1, 5, 6 and 10 had a greater number of
alleles when compared to TLR2, 3, 4, 7, and 9, which is in line with the previous works
(Georgel et al., 2009). Finally, balancing selection was also detected on TLR1, 6 and 10
(Ferrer-Admetlla et al., 2008). These divergent results are elucidating the differences among
populations.
Box 2. Types of natural selection. (Figure adapted from Quintana-Murci 2016)
32
Thus, variability of the TLRs should be assessed within the same host species, or even breed,
because it has been seen that the type of selection and the number of genetic variants differed
on certain populations from different geography (Ferrer-Admetlla et al., 2008)(Barreiro et al.,
2009); and TLRs responses differ in a species-specific manner (Werling et al., 2009).
In Chapter 3.1, we describe the genetic polymorphisms of TLRs on canids. Both adaptive and
non-adaptive evolution should be considered in dogs. Non-adaptive evolution has probably a
more significant role than in other species due to dogs were under a first bottleneck with
domestication and a second one with the artificial selection of breeds (Figure 15) (Lindblad-
Toh et al., 2005). For this reason it should be taken into account the need for dealing with
different breeds, and even with other wild canids such as the wolf for the analysis of canine
TLR polymorphism.
Figure 15. Evolution forces acting on TLR genes. Adaptive evolution caused by infections, combined with
non-adaptive evolution caused by genetic drift, population bottlenecks and migration routes, contribute to TLR
polymorphisms. Some polymorphisms can lead to resistance to infections, and other to autoimmunity
phenomena. (Figure adapted from Netea, Wijmenga, and O’Neill 2012).
1.2.3.1.2. TLRs in skin diseases
Toll-like receptors are innate immune sensors that have a role in both identifying commensal
microbiota and maintaining the homeostasis, and in disease establishment. On one hand,
TLRs promote mutually beneficial commensal-host interactions (Kubinak and Round, 2012).
On the other hand, they are responsible for some diseases, including cutaneous ones (Table
2).
33
TLRs are expressed not only at innate immune cells, but also at skin cells such as
keratinocytes, mast cells, stromal cells and adipocytes (Nestle et al., 2009). Some relationships
between skin diseases and TLRs have been described. Thus, for specific skin diseases either
certain TLR polymorphisms are correlated to increased susceptibility or resistance or TLR
expression is increased or decreased (Table 2). Due to this clear link TLR-disease, therapies
targeting TLRs are being assessed for dermatological diseases, using for example molecules
that have the ability to modulate TLR expression (TLR agonists and antagonists) (Matin et al.,
2015).
Specifically on dog skin, only Leishmania infected dogs have reported an altered expression of
TLR2, 3, 4 and 9 (Figueiredo et al., 2013; Hosein et al., 2015; Melo et al., 2014). When
assessing other canine diseases some links have been reported in both differential TLR
expression and polymorphisms.
In gastrointestinal pathologies, differential expression of TLR2 has been described in
Inflammatory Bowel Disease (IBD) (McMahon et al., 2010); TLR2, 4, 5 and 9 in chronic
enteropathies in German Shepherd (Burgener et al., 2008)(Allenspach et al., 2010); and TLR2
and 4 in inflammatory colorectal polyps (Yokoyama et al., 2017). Moreover, genetic
polymorphisms in TLR4 and TLR5 have been associated with IBD in German Shepherd dogs
(Kathrani et al., 2010), but only protective SNPs from TLR5 have been associated with IBD in
other 38 dog breeds (Kathrani et al., 2011). Other pathologies presented differential
expression of TLRs, such as TLR4 in osteoarthritis (Kuroki et al., 2010) and in infected canine
endometrium (Chotimanukul and Sirivaidyapong, 2011); TLR2 and 7 in arthritis (Riggio et al.,
2014); and TLRs 1-4, 6-10 in sino-nasal aspergillosis and idiopathic lymphoplasmacytic rhinitis
(Mercier et al., 2012).
34
Table 2. Dermatological diseases associated with an alteration on Toll-like Receptors.
Skin disease TLRs alterations References
Infectious diseases
Acne vulgaris Increased expression of TLR2, correlated with severity (Bakry et al., 2014; Kim et al., 2002)
Candidiasis TLR9-deficient mice resistant to candidiasis (Kasperkovitz et al., 2011)
Leishmaniosis Increased expression of TLRs 2, 4 and 9 (Tuon et al., 2010)
Leprosy Hypo-functional TLR1 variant protected against leprosy (Wong et al., 2010)
Polymorphisms on TLRs 1, 2 and 4 associated with susceptibility (Bochud et al., 2009)
Increased expression of TLR1 and 2 (Krutzik et al., 2003)
Rosacea Increased expression of TLR2 (Yamasaki et al., 2011)
Staphyloccocal infections Increased expression of TLR2 (Hilmi et al., 2014)
Immunological diseases
Atopic dermatitis Polymorphisms on TLR2 associated with the disease (Potaczek et al., 2011)
Decreased expression of TLR2 and 4 (Lesiak et al., 2012)
Psoriasis Increased expression of TLR1, 2, 4, 5, 9 (Begon et al., 2007)
Lichen planus Increased expression of TLR4 and reduced expression of TLR2 (Janardhanam et al., 2012)
Increased expression of TLR4 and 9 (Siponen et al., 2012)
Decreased expression of TLR1 and 2 (Salem et al., 2013)
Pemphigus Increased expression of TLRs 2, 3 and 4 (Abida et al., 2013)
Systemic autoimmune diseases (with skin manifestations)
Systemic Lupus Erythematosus Pathogenic role for TLR7 and protective role for TLR9 (Celhar et al., 2012)
Sarcoidosis Increased expression of TLRs 2, 3, 4, 5, 6, 7, and 8 (Huizenga et al., 2015)
Systemic sclerosis Increased expression of TLR3 (Agarwal et al., 2011; Farina et al., 2010)
Cancer
Basal cell carcinoma Higher expression of TLRs 1, 2, 3, 5, 6, 7, 8 (Muehleisen et al., 2012)
Squamous cell carcinoma Higher expression of TLRs 1, 2, 3, 5, 6, 7, 8 (Muehleisen et al., 2012)
Other skin diseases
Stevens–Johnson syndrome Polymorphisms on TLR3 associated with the disease (Ueta et al., 2007)
35
1.3. Skin microbiota
The microbiota is defined as a collection of microorganisms that inhabit a specific
environment (Marchesi and Ravel, 2015), thus the canine skin microbiota is the collection of
microorganisms that inhabits the skin of dogs.
Until today (June 2017), 637 out of 846 studies found in pubmed regarding skin microbiota are
human-based, so we will provide an extensive overview of skin microbiota focusing first on
what is known for human and then specifically for dog in health (Chapter 1.3.1, 1.3.2 and
1.3.3) and disease (Chapter 1.3.4). We will finish this chapter reviewing the clinical potential
of the microbiota (Chapter 1.3.5).
36
1.3.1. What is living on healthy skin?
The human skin microbiota is predominantly inhabited by bacteria, followed by fungi and
viruses with lower abundances (Oh et al., 2014). Microbiota members mostly inhabit the skin
surface, hair follicles and other appendages (Figure 16)(Grice and Segre 2011). Moreover,
Nakatsuji and colleagues also detected bacteria in the deep dermis and subcutaneous tissues of
healthy individuals (Nakatsuji et al., 2013).
Figure 16. Human skin microhabitats and their associated microbiota. Human skin has three main microhabitats: dry, moist and sebaceous. Each of them has its different skin appendages and specific physicochemical properties that define the microbiota they harbor (Figure from Barnard and Li 2017).
First human microbiota studies using next-generation sequencing techniques elucidated in skin
that the three main microhabitats harbor specific bacteria creating bacterial signatures. So,
sebaceous sites are inhabited by Propionibacterium spp; moist sites, by Staphylococcus and
Corynebacterium spp; and dry sites with gram-negative microorganisms (Grice et al. 2009;
Costello et al. 2009; Grice and Segre 2011) (Figure 17a). Dry skin sites present higher diversity
values when compared to the others. Similarly, when compared to other body site
microbiotas, skin presents higher diversity values despite the lower microbial load (Belkaid and
Segre, 2014a).
Dog skin is almost entirely covered by a dense fur creating a more uniform
microenvironment, so no clearly defined microhabitats have been identified. Bacteria from the
Proteobacteria phylum are the most abundant all over the dog skin (Figure 17b) ( Song et al.
2013; Rodrigues Hoffmann et al. 2014), contrasting with the microhabitat-specific taxa on
human. Within the skin sites, haired skin regions usually present higher alpha diversity values
when compared to mucosal or mucocutaneous regions (Rodrigues Hoffmann et al., 2014).
37
Besides skin site divergences, a large diversity between individuals has also been reported
(Hoffmann et al., 2014).
Figure 17. Skin microbiota composition per skin site. In a) human skin microbiota classified by microhabitat (Figure from Grice and Segre 2011) and in b) dog skin microbiota (Figure from Rodrigues Hoffmann et al. 2014).
38
Human skin microbiota presents high variability, not only between skin sites or microhabitats
–intra-individual variability–, but also between individuals –inter-individual variability (Grice
and Segre, 2011; Human Microbiome Project Consortium, 2012; Oh et al., 2014).
Individual signatures of the skin microbiota have the ability to identify items that an individual
came in contact with (Fierer et al., 2010; Lax et al., 2015; Meadow et al., 2014). Despite the
temporal stability is a personalized feature (Flores et al., 2014; Oh et al., 2016), skin microbiota
in sebaceous and dry sites is relatively constant over time (Figure 18) (Oh et al., 2016). Finally
strain-level differences in Propionibacterium acnes were individual-specific, whereas those in
Staphylococcus epidermidis were more site-specific (Oh et al., 2014). For all these reasons
demonstrating individual signatures in skin, microbiota analyses have even been proposed as a
new tool for forensic science (Hampton-Marcell et al., 2017).
Figure 18. Temporal stability of the human skin microbiota. Healthy adults maintain their skin microbial communities at sebaceous and dry sites over time despite the constant exposure to external environment. (Excerpted from Oh et al. 2016).
Microbial diversity is not limited to bacteria; microorganisms such as archaea, fungi, and
viruses also have major roles in human health and disease (Eckburg et al., 2003; Handley,
2016; Peleg et al., 2010) and whole genome sequencing approaches have allowed deciphering
this global picture. Depending on the habitat, the abundance of fungi and viruses varies from
<0.1% of microorganisms in the gastrointestinal tract to up to 10% and 40% on skin,
respectively (Oh et al. 2014; Belkaid and Segre 2014). Archaea can be found on
gastrointestinal tract with <1% (Arumugam et al., 2011) and are nearly absent on skin (Oh et
al., 2014).
39
Figure 19. Microbial community of the skin using shotgun metagenomics. Relative abundances of viral, bacterial, and fungal components per skin site. Sites represent three skin microhabitats: sebaceous (blue), dry (red), and moist (green) (excerpted from Belkaid and Segre 2014)
Fungal microbiota can also be retrieved using amplicon-based approaches targeting their
rRNA gene to amplify the Internal Transcribed Spacer (ITS) regions or 18S (Halwachs et al.,
2017).
Malassezia species were detected as the main colonizers of the human skin in forearm and face
(Paulino et al., 2006; Zhang et al., 2011). Findley and colleagues found that in fact most of the
skin sites in healthy individuals were colonized by Malassezia, detecting site signatures only at
species level. By contrast, the three foot sites analyzed presented more diversity being
colonized by a combination of Malassezia, Aspergillus, Cryptococcus, Rhodotorula, Epicoccum, and
others (Findley et al., 2013). This pattern seemed to be specific to adult population, whereas
pre-pubertal skin presented higher diversity (Jo et al., 2016). When analyzed with
metagenomics approaches, fungi presented high abundances near the ears and forehead, but
low representation in the feet (<1%) despite the high diversity observed in the previous
amplicon-based studies (Oh et al., 2014).
On dog skin using an amplicon-based approach, Meason-Smith and colleagues found that the
most abundant fungi were Alternaria and Cladosporium independently from the skin site, which
was not an influencing factor on mycobiota composition and structure (Figure 20). In fact,
different skin sites tended to be similar within a dog (Meason-smith et al., 2015). Alternaria and
Cladosporium were also the most abundant fungi on cat skin mycobiota (Meason-Smith et al.,
2017).
40
Figure 20. Skin mycobiota composition on healthy dogs per skin site. (Excerpted from Meason-smith et al. 2015).
The viruses are the most forgotten microbiota members, and main insights on human skin
come from whole shotgun metagenomics approaches (Foulongne et al., 2012; Oh et al., 2014,
2016; Wylie et al., 2012). Only one study enriched viral particles and assessed cross-sectional
diversity within different skin sites, which found that skin virome was highly site-specific and
modulated by occlusion and exposure, in addition to sebum and moisture (Hannigan et al.,
2015).
The virome can be classified within two main groups: prokaryotic virome, which includes
viruses affecting prokaryotes, such as bacteriophages; and eukaryotic virome, which includes
viruses affecting eukaryotic cells (Virgin, 2014). Among the prokaryotic virome the most
abundant members were Propionobacterium and Staphylococcus phages (Hannigan et al., 2015; Oh
et al., 2014, 2016) and Pseudomonas and Bacillus phages (Hannigan et al., 2015). In eukaryotic
skin virome, Papillomaviruses as well as Polyomavirus were the most commonly found (Foulongne
et al., 2012; Oh et al., 2014, 2016; Wylie et al., 2012); however the most abundant viruses were
specific to single individuals (Oh et al., 2016). On dog skin, the virome has not been assessed
yet.
41
1.3.2. Factors shaping bacterial skin microbiota
The key variables driving human skin microbiota structure and composition are skin site
followed by the individual, as seen in the previous section. Besides, many other variables are
affecting its composition and structure, which can be globally classified as: intrinsic factors (or
host-specific) and extrinsic factors (or environment-specific) (Grice and Segre 2011).
On one hand, host factors include those relative to the individual such as the age, gender,
racial origin or immune system and all of them seem to have an effect on skin microbiota
(Sanmiguel and Grice, 2015).
Regarding the age, skin microbiota evolves over the first year of life, showing an initial
colonization by Staphylococci that decreases with age as diversity increases (Capone et al., 2011).
Moreover, progressive microbial shifts in skin and nares have been detected in different sexual
maturation stages and when comparing children to adults (Oh et al., 2012). Cohorts usually
include people from both genders, but only few studies have found a strong link between
gender and skin microbiota structure and composition in hands (Fierer et al., 2008), upper
buttock (Zeeuwen et al., 2012) and axillary vault (Callewaert et al., 2013). Racial origin or
ethnicity presented a strong effect on skin microbiota when linked to clinical metadata, as it
was shown by Human Microbiome Project that included around 300 individuals from
different genetic backgrounds (Asian, Black, Mexican, Puerto Rican and White) (Human
Microbiome Project Consortium, 2012). Other studies detected differences in skin microbiota
of Chinese people when compared to other racial groups (Leung, Wilkins, and Lee 2015).
Also, a recent study including six different ethnic groups living in New York City detected
racial origin as a secondary factor shaping skin microbiota after skin site (Perez Perez et al.,
2016). Finally, the immune system is the main responsible of recognizing the microorganisms
and it definitely affects skin microbiota, as well as skin microbiota affects immunity (Naik et
al. 2012; Naik et al. 2015). This part is covered in the following section (Chapter 1.3.3).
On the other hand, extrinsic or environmental factors include those derived from the
surroundings (geography, urbanization or cohabitation) or personal habits (hygiene, cosmetics
use or diet).
Tribal populations that spend more time outdoors presented higher abundances of
environmental-derived taxa and increased alpha diversity values when compared to other
populations (Clemente et al., 2015; Hospodsky et al., 2014). Skin microbiota varied also when
comparing different environments of a delimitated region (Hanski et al., 2012; Ying et al.,
2015). In line with this, individuals cohabiting together –even when including pets– shared a
larger proportion of skin microbiota if compared to other individuals (Song et al. 2013; Misic
et al. 2015). Regarding personal habits, Fierer and colleagues found that skin hand microbiota
was influenced not only by gender, but also by the time since the last hand washing. In fact,
the gender differences they reported could also have been due to extrinsic factors such as the
use of cosmetics or moisturizers (Fierer et al., 2008). A recent study found that soap and
shampoo practices were secondary factors shaping skin microbiota on New Yorkers (Perez
Perez et al., 2016).
42
Finally some of these microbial drivers are both intrinsic and extrinsic (racial origin – different
environment; or sex – different hygiene practices), and even some of them are a combination
of both. For example, Lehtimäki and colleagues recently reported that geography shapes
microbiota in an age-dependent way: environmental effects decrease as age increase
(Lehtimäki et al., 2017).
In canine skin, all of these factors have the potential to drive skin microbiota composition and
structure. Moreover, those derived from the environment are likely having a major impact in
dogs than in humans. On one hand, as host factors we could include sex, age, skin site,
individual, immune system, genetic background or even hair coat. On the other hand, as
environmental factors we could include cohabitation, lifestyle, environment, season, diet,
hygiene or geography (Figure 21).
Figure 21. Canine skin microbiota is potentially shaped by both host and environmental factors.
Few of these factors have been tested as potential drivers of skin microbiota structure and
composition in healthy dogs.
Rodrigues-Hoffmann and colleagues described the skin site effect on dog skin microbiota,
showing mucosal and muco-cutaneous regions less diversity when compared to haired skin
sites. They also detected an individual effect, despite not assessing it directly. Other factors
such as the presence of fleas, time spend outdoors vs indoors, the sex and age were assessed as
potential drivers of canine skin microbiota without finding any relationship. This could have
been due to they worked with a small (N=12) and heterogeneous cohort including dogs from
different households, genetic backgrounds and ages, so true effect could have been obscured
(Rodrigues Hoffmann et al., 2014). Besides, animals cohabiting together (similar to humans)
share more skin microbiota and moreover cohabitation was affecting skin microbiota of the
pet owner, increasing its bacterial diversity (Song et al. 2013).
43
1.3.3. Commensal microbiota functions on the skin
Skin is one of the main interfaces with the environment and it is constantly exposed to
external microorganisms, which some of them can be potential pathogens. Besides that, skin is
not only exposed but also colonized by many microorganisms that constitute the skin
microbiota.
This living barrier that is the skin microbiota contributes to the skin immunity mainly in three
different ways: directly inhibiting pathogen growth; enhancing host innate immunity; and
educating and priming adaptive immunity (Figure 22) (Sanford and Gallo, 2013). We will not
review how the skin microbiota shapes the adaptive immunity, as it is beyond the scope of this
dissertation.
Figure 22. Skin microbiota immune functions on health. Host cells and skin microbiota are associated and
their cross-talks help skin immunity through several mechanisms. (Excerpted from Sanford and Gallo 2013).
Skin microbiota members are inter-communicated (bacteria-bacteria interactions) and
moreover interact with host cells establishing cross-talks (bacteria-host cells interactions).
These interactions can be classified in three main categories: commensalism, when one
benefits from the other; mutualism, when both benefit; and detrimental relationship, when
one harms the other (Figure 23) (Schommer and Gallo, 2013).
44
Figure 23. Dynamics of microbial interaction at the skin surface. Members of the microbiota form complex interaction networks, with both microbe-microbe and microbe-host interactions. (Excerpted from Schommer and Gallo 2013)..
Usually these interactions and their functions are studied in individual commensal members of
the microbiota. Among all the bacteria inhabiting skin microbiota, we delve into the functions
of the two most well-characterized microbiota members on human skin: Propionibacterium acnes
and Staphylococcus epidermidis (Table 3).
Table 3. Propionibacterium acnes and Staphylococcus epidermidis contributions to the innate
immunity functions.
Commensal bacteria Functions in skin health
Propionibacterium acnes Producing SCFAs that inhibit growth of other microorganisms
Maintaining acidic skin pH
Producing bacteroicins
Promoting commensal growth
Staphylococcus epidermidis Producing AMPs and bacteroicins
Promoting host immune to produce host AMP
Promoting host immune responses via TLR signaling
Propionibacterium acnes mechanisms that contribute to skin immunity are mostly based on
directly inhibiting pathogen growth, which is achieved by: 1) metabolizing triglycerides from
the sebum to short chain fatty acids, which have antimicrobial properties and contribute to
acidic pH of the skin (Shu et al., 2013; Ushijima et al., 1984); and 2) producing bacteroicins
(Faye et al., 2011), which are toxins that inhibit the growth of similar bacteria. Moreover, P.
acnes strains associated with healthy skin –not those ones associated with acne–, carry genes
that synthesize thiopeptides, which are antimicrobial compounds that inhibit the growth of
gram-positive species (Christensen and Brüggemann, 2014).
45
Staphylococcus epidermidis contributes to skin immunity by both enhancing and modulating innate
immunity via TLR signaling (Lai et al., 2009; Wanke et al., 2011) and by directly inhibiting
pathogen growth producing antimicrobial peptides (Bastos et al., 2009; Cogen et al., 2010).
For example, Staphylococcus epidermidis was able to amplify the innate immune response against
the invading pathogen Staphylococcus aureus by increasing antimicrobial peptides expression and
abolishing the inhibition of NF-κB signaling asserted by the pathogen (Figure 24) (Wanke et
al., 2011).
Figure 24. Staphylococcus epidermidis cross-talk with innate immunity. Staphylococcus epidermidis produces antimicrobial peptides to act as a barrier against colonization of potentially pathogens. In addition, it secretes a small molecule, which increases expression of defensins through TLR2 signaling. (Excerpted from Gallo and Nakatsuji 2011).
46
1.3.4. Skin microbiota and dermatological diseases
In a healthy skin, homeostasis is maintained through a cross-talk between immune system and
microbiota and when dysregulated produces inflammation (Belkaid and Segre 2014). In
humans, several studies have associated skin diseases to dysbiosis, which is an imbalance of
microbiota (Table 4 and Figure 25). However, if dysbiosis is the cause or the consequence of
the disease in most cases still remains to be clarified. On dogs, only few studies have been
performed and are focused on canine atopic dermatitis (Table 4).
Figure 25. Dermatological diseases associated with dysbiosis on skin microbiota. (Excerpted from (Barnard and Li, 2017).
47
Table 4. Dermatological diseases associated with skin microbiota alterations.
Disease Skin microbiota alterations References
Acne vulgaris Strain-level differences in P. acnes are associated with acne. (Fitz-Gibbon et al., 2013)
Balance between acne- and health-associated species determines health status. (Barnard and Li, 2017)
Increased Staphylococci, correlated with severity. (Dreno et al., 2017)
Atopic dermatitis (AD) Increased Staphylococcus, correlated with severity. Decreased diversity during flares. (Kong et al., 2012a)
Increased Staphylococcus and Corynebacterium, correlated with severity (in AD and PID). (Oh et al., 2013)
Differences btw affected vs unaffected skin. Stenotrophomonas prevalent in therapy responders. (Seite and Bieber, 2015)
1-year-old infants with AD were not colonized with S. aureus before developing AD. (Kennedy et al., 2017)
Signature in AD-prone skin: enriched in Streptococcus and Gemella but depleted in Dermacoccus. (Chng et al., 2016)
Bacterial infection (H. ducreyi) Resolvers and pustule formers have distinct skin bacterial communities (Rensburg et al., 2015)
Primary immunodeficency Decreased site specificity and temporal stability. Colonization with species not found in controls. (Oh et al., 2013)
Increased Gram-negative bacteria, especially Acinetobacter. Decreased Corynebacterium. (Smeekens et al., 2013)
Psoriasis vulgaris Increased Firmicutes and Actinobacteria. (Gao et al., 2008)
Decreased Staphylococci and Propionibacteria. (Fahlén et al., 2012)
Increased Corynebacterium, Propionibacterium, Staphylococcus, and Streptococcus. (Alekseyenko et al., 2013)
Vitiligo Decreased diversity. Altered networks: Firmicutes as central nodes (Actinobacteria in health) (Ganju et al., 2016)
Wounds (non-chronic) Troughout healing process, wound microbiota became increasingly similar to adjacent skin. (Hannigan et al., 2014)
Wounds (chronic) Prevalent anaerobic bacteria (from Clostridiales family XI) (Price et al., 2009)
Prevalent uncharacterized Bacteroidales, anaerobes and Staphylococcus, Corynebacterium, and Serratia (Wolcott et al., 2009)
Increased anaerobic bacteria and Corynebacterium and Staphylococcus. (Gontcharova et al., 2010)
Canine atopic dermatitis Decreased species richness on allergic dogs. (Rodrigues Hoffmann et al., 2014)
Decreased bacterial diversity, increased Staphylococcus (S. pseudintermedius) and Corynebacterium. (Bradley et al., 2016a)
Increased abundance of S. pseudintermedius at the site of lesion induction. (Pierezan et al., 2016)
48
In most of the previous studies, affected skin showed lower diversity when compared to
healthy one, independently from the skin pathology. This lower diversity was usually linked
to an overgrowth of specific taxa, which sometimes was also correlated to severity.
Staphylococcus is a genus that was overrepresented in acne vulgaris (Dreno et al., 2017),
atopic dermatitis (Kong et al. 2012; Oh et al. 2013), primary immunodeficiencies (Oh et al.,
2013), psoriasis (Alekseyenko et al., 2013) and chronic wounds (Gontcharova et al., 2010;
Wolcott et al., 2009) and it was correlated with severity for acne (Dreno et al., 2017), atopic
dermatitis (Kong et al. 2012; Oh et al. 2013) and primary immunodeficiencies (Oh et al.,
2013). However, another study in Psoriasis detected a decrease on this genus (Fahlén et al.,
2012). Corynebacterium was also a common genus altered in several pathologies: increased in
atopic dermatitis and primary immunodeficiencies (Oh et al., 2013), psoriasis (Alekseyenko
et al., 2013), and chronic wounds (Gontcharova et al., 2010; Wolcott et al., 2009); and also
decreased in primary immunodeficiencies (Smeekens et al., 2013). Besides these highly
common trends, each skin disease presented its own microbial characteristics and even
some studies lead to contradictory results (Table 4).
In dogs, the three studies targeting atopic dermatitis found also that affected skin presented
lower diversity when compared to healthy one (Bradley et al., 2016a; Pierezan et al., 2016;
Rodrigues Hoffmann et al., 2014). Staphylococcus pseudointermedius increased in affected skin,
either when compared to healthy dogs (Bradley et al. 2016) or the same dog in a non-
affected site (Pierezan et al., 2016). Also Corynebacterium increased in affected skin (Bradley
et al., 2016).
1.3.5. Skin microbiota as a clinical tool
As skin microbiota differs in health and disease, some researchers have suggested that
microbiota analyses have the potential to be used in clinics as a diagnosis, prognosis and
even therapeutic tool (Grice 2014; Grice 2015).
All these potential clinical approaches are based on microbiota property to sense its
environment and respond against alterations. These alterations can be harmful or
beneficial. They are harmful when they create a dysbiotic state, which can be produced by
pathogens, environmental stimuli or immune defects among others. Beneficial alterations
are usually intentioned and aim to recover the homeostasis, which can be achieved using
products such as pre- and pro-biotics or even antibiotics (Egert and Simmering, 2016).
As previously seen, some skin diseases have a kind of bacterial signature, with some
overrepresentation of specific bacteria (Table 4). Moreover, as seen in acne vulgaris,
sometimes a skin disease is not due to the presence of a specific set of bacteria but to an
altered balance between health- and disease-specific bacteria (Barnard and Li, 2017). So,
microbiota analysis could be used as a diagnosis tool for certain skin diseases identifying
the microbial balance and signature.
Regarding prognosis, higher proportions of specific bacteria are associated to increased
severity of certain dermatological diseases, such as atopic dermatitis or acne vulgaris
(Dreno et al., 2017; Kong et al., 2012b; Oh et al., 2013). Moreover, certain skin microbiota
49
compositions make you a resolver or a pustule former in an infection with Haemophylus
ducreyi (Rensburg et al., 2015) and also people prone to atopic dermatitis have a different
skin microbiota when compared to healthy individuals (Chng et al., 2016). In non-chronic
wounds, a specific microbiota profile was correlated with future complication (Hannigan et
al., 2014). Thus, different researchers have confirmed the potential to use the microbiota as
a prognosis tool in several skin conditions.
Finally, the most promising tool in clinics is using microbiota manipulation as a therapeutic
approach to resolve certain diseases (Grice, 2014; Reid et al., 2011). These manipulations
can be divided in two main groups: the classical antimicrobial approach, to reduce the
number of pathogenic microorganisms; and the novel pre- and pro-biotic approach, to
increase a specific subset of beneficial bacteria (Egert and Simmering, 2016). Despite the
function of antibiotics is to restore the healthy status through pathogen elimination, it also
affects microbiota and has detrimental effects such as killing beneficial microorganisms and
creating antibiotic resistances with its long-term use. Moreover, antibiotics perturb the
original microbiota and the altered version is more likely to present overgrowth of
opportunistic pathogens (for a review on microbiota alterations due to antibiotics use, see
Ferrer et al. 2017). All these concerns about antibiotics use led researchers to investigate
the novel approach of using pre- and probiotics as a therapy or complement for certain
diseases.
Prebiotics are non-digestible food ingredient that affects the host by stimulating the growth
or activity of specific beneficial bacteria, whereas probiotics are products containing living
microorganisms in sufficient numbers, so they can alter the microbiota and produce
beneficial health effects in the host (Schrezenmeir and de Vrese, 2001) (Fig 26A). When
developing pre- and pro-biotics, researchers need to target specific bacterial products that
are beneficial for the skin (Fig 26B) and therefore find the bacterial strain that metabolizes
that product or the substrate that promotes the growth of that specific strain (Lew and
Liong, 2013). Microbiota studies are invaluable to understand the mechanisms and
dynamics of these novel products and to ensure their safety. Most of pre- and pro-biotics
aim to manipulate the gastrointestinal microbiota rather than the skin, even when treating
dermatological problems (Baquerizo Nole et al., 2014; Krutmann, 2009). Some topical
probiotics have shown an effect until the date: Lactobacillus rhamnosus for atopic dermatitis
(Hoang et al., 2010; Viljanen et al., 2005) ; Bifidobacterium longum for sensitive skin (Guéniche
et al., 2009); and kefir –uncharacterized probiotic mixture– for wound healing (Huseini et
al., 2012). Also sphingomyelinase produced by Streptococcus thermophilus was shown to
increase skin-ceramide levels in aged subjects (Dimarzio et al., 2008).
50
Figure 26. Pre- and probiotics for the skin. a) An example to illustrate the concepts of pre- and pro-biotics for skin and b) bacterial products with beneficial effects on the skin (Figures from (Egert and Simmering, 2016) (a) and (Lew and Liong, 2013) (b)).
51
When speaking about microbiota used as a therapy, we must include microbiota
transplantation from healthy donors to affected receptor. Fecal transplants have already
been used in human to treat Clostridium difficile recurrent infection since 40 years ago with a
success rate up to 92%-100% (Gough et al., 2011), but the underlying microbiota dynamics
are just beginning to be understood (Song et al. 2013; Fuentes et al. 2014; Leber et al. 2015;
Seekatz et al. 2016; Staley et al. 2017). On the skin of a murine model for atopic dermatitis,
Myles and colleagues tried some kind of “microbiota transplantation” –only culturable
gram-negative bacteria– that led the mice to an improved outcome (Myles et al., 2016).
Most pre- and probiotics products worked without understanding its dynamics and mainly
in the gut. Networks dynamics aim to give light in these mechanisms, and are classified in
three types: individual dynamics; group dynamics; and universal dynamics (Figure 27)
(Bashan et al., 2016). Nowadays, network studies on microbiota dynamics of healthy
individuals elucidated that some body sites presented universal dynamics (similar among
the individuals), whereas others were highly personalized (Bashan et al., 2016). In fact, the
healthy gut microbiota presented universal dynamics, which could be the reason why most
of the pre- and pro-biotics work on this body site. In contrast, among the several skin sites
included in this study only the retroauricular crease presented universal dynamics, whereas
forehead, palm, and antecubital fossa presented individual dynamics. Thus, generic
microbiota manipulations may result ineffective or even detrimental in some body sites
(Bashan et al., 2016).
52
Figure 27. Alternative scenarios of microbial dynamics across different healthy individuals. Nodes represent microbial species and edges represent interspecies interactions (green and red edges represent co-ocurrence or mutual exclusion interactions, respectively). The three main scenarios are: a) individual dynamics, where the underlying dynamics is unique for each individual; b) group dynamics, where dynamics are shared across different groups; and c) universal dynamics, where different subjects have the same underlying dynamics. (Excerpted from Bashan et al. 2016)
To conclude, microbiota analyses have the potential to diagnose, prognose and even treat
several skin diseases. To design safe products and accurate diagnoses and prognoses tests,
healthy and altered microbiota should be better defined, as well as microbial dynamics
should be fully understood.
53
2. Objectives
The overall goal of this thesis is to expand the knowledge on what defines the healthy
status of skin in dogs, first at innate immunity level but mainly characterizing skin
microbiota. Moreover, this is an industrial PhD project in collaboration with Vetgenomics,
SL., so this knowledge will provide the scientific background and the know-how needed to
develop a new strategic business area on Vetgenomics focused on microbiota analyses.
To provide insights on dog innate immune system (Chapter 3.1), we aim to:
- Design and validate an individual genotyping array of non-synonymous
polymorphisms on canine Toll-like Receptor (TLR) genes; creating a tool to
characterize the innate immune system at TLR level.
- Describe the normal variability of TLR genes on different cohorts of healthy dogs
and wolves.
To provide insights on skin microbiota in healthy dogs, we aim to:
- Assess the effect of breed, individual or skin site on shaping skin microbiota in a
pilot and heterogeneous cohort (Chapter 3.2).
- Assess the effect of either host or environmental factors (such as skin site, surgery,
sex, date of birth date, pigmentation, and time spent in the kennel) on shaping skin
microbiota (Chapter 3.3).
- Finally, we aim to assess the potential of the 3rd generation single-molecule
sequencing technology for microbiota studies using Nanopore sequencing (Chapter
3.4).
54
55
3. Results
This section, divided in four chapters, encompasses a description of the results obtained in
four different studies I carried out during the last three years.
Chapter 3.1 describes the genetic variability in canine TLR genes focusing on non-
synonymous mutations, by massive sequencing and individual genotyping.
Chapter 3.2 depicts the skin microbiota structure and composition on a heterogeneous
cohort of nine healthy dogs, from three different breeds inhabiting different households.
Chapter 3.3 depicts the skin microbiota structure and composition on a homogeneous
cohort of thirty-five healthy dogs, from the same cross-breed sharing a common
environment.
Chapter 3.4 assesses the potential of sequencing the full-length 16S rRNA gene with
MinION™ (3rd generation sequencer) to perform a microbiota study.
56
3.1. Non-synonymous genetic variation in
exonic regions of canine Toll-like receptors
This chapter consists of the article entitled “Non-synonymous genetic variation in exonic
regions of canine Toll-like receptors” published in Canine Genetics and Epidemiology
journal in October 2014 (1:1).
Supplementary material of this article is available at Canine Genetics and Epidemiology
online: https://cgejournal.biomedcentral.com/articles/10.1186/2052-6687-1-11
RESEARCH Open Access
Non-synonymous genetic variation in exonicregions of canine Toll-like receptorsAnna Cuscó1,2, Armand Sánchez1, Laura Altet2, Lluís Ferrer3 and Olga Francino1*
Abstract
Background: Toll-like receptors (TLRs) are pattern recognition receptors (PRRs) considered to be the primarysensors of pathogens in innate immunity. Genetic variants could be associated to differences in breed innateimmune response to pathogens and thus to susceptibility to infections or autoimmune diseases. There is thereforegreat interest in the characterization of canine TLRs.
Results: Polymorphisms in canine TLRs have been characterized by massive sequencing after enrichment of theirexonic regions. DNAs from 335 dogs (seven different breeds) and 100 wolves (two different populations) were usedin pools. The ratio of SNP discovery was 76.5% (in relation to CanFam 3.1); 155 out of 204 variants identified werenew. Functional annotation identified 64 non-synonymous variants (43 new), 73 synonymous variants (56 new) and67 modifier variants (57 new). 12 out of 64 non-synonymous variants are breed or wolf specific. TLR5 has beenfound to be the most polymorphic among canine TLRs. Finally, a TaqMan OpenArray® plate containing 64 SNPswith a possible functional effect in the protein (4 frameshifts and 60 non-synonymous codons) has been designedand validated.
Conclusions: Non-synonymous genetic variation has been characterized in exonic regions of canine Toll-like Receptors.The TaqMan OpenArray® plate developed to capture the individual variability that affects protein function will allowhigh-throughput genotyping either to study association to infection susceptibility or even TLR evolution in the caninegenome.
Keywords: TLRs, Toll-like receptor, Polymorphism, SNPs, Non-synonymous SNPs, Canine, Dog, Innate immunity
Lay summaryToll-like receptors (TLRs) are pattern recognition recep-tors (PRRs) and are the primary sensors of pathogens inthe body. Genetic variants could be associated with differ-ences in breed response to pathogens and also to suscepti-bility to infections and/or autoimmune diseases. There isgreat interest in the characterization of canine TLRs.Genetic variation in canine TLRs has been characterized
using massive parallel sequencing. DNA from 335 dogs(seven breeds: Beagle, German Shepherd dog, Yorkshireterrier, French bulldog, Boxer, Labrador and Shar Pei) plus100 wolves (two populations: Iberian and Russian) weresequenced in 16 pools of 25 dogs or 50 wolves. In total,we found 204 variants, of which 155 were new. Compa-rison of these variants with the published dog genome
sequence (called CanFam 3.1) Functional annotation iden-tified 64 non-synonymous variants (43 new), 73 synonym-ous variants (56 new) and 67 modifier variants (57 new).Twelve of 64 non-synonymous variants were breed or wolfspecific. TLR5 has been found to be the most polymorphicamong canine TLRs. Finally, a TaqMan OpenArray(R)plate containing 64 SNPs with a possible functional effectin the protein (4 frameshifts and 60 non-synonymouscodons) has been designed and validated.Non-synonymous genetic variation has been character-
ized in exonic regions of canine Toll-like Receptors. TheTaqMan OpenArray(R) plate developed to capture the in-dividual variability that affects protein function will allowhigh-throughput genotyping either to study association toinfection susceptibility or even TLR evolution in the caninegenome.
* Correspondence: Olga.Francino@uab.cat1Molecular Genetics Veterinary Service. Veterinary School, UniversitatAutònoma de Barcelona, Barcelona, SpainFull list of author information is available at the end of the article
© 2014 Cuscó et al.; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the CreativeCommons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, andreproduction in any medium, provided the original work is properly credited. The Creative Commons Public DomainDedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article,unless otherwise stated.
Cuscó et al. Canine Genetics and Epidemiology 2014, 1:11http://www.cgejournal.org/content/1/1/11
BackgroundToll-like receptors (TLRs) are the most widely studiedpattern recognition receptors (PRRs) and are consideredto be the primary sensors of pathogens in innate im-munity. These molecules are constituted by leucine-richrepeat (LRR) domains, a unique intramembrane domainand a Toll/Interleukin-1 receptor (TIR) domain. Pathogen-associated molecular Patterns (PAMPs) are sensed throughLRR domain, and signals are transduced through TIRdomain, which is always located in the cytoplasm, in orderto activate innate immunity response (for a review, see [1]).Ten TLRs have been identified in dogs. TLRs can be
classified into groups, depending on the PAMPs detectedand their cellular location. TLR 1, 2, 4, 5 and 6 detectpathogen extracellular components. TLRs 3, 7, 8 and 9target nucleic acids. The ligand for TLR10 is unknown [2].Another way to classify TLRs is their cellular location.
TLRs 1, 5, 6 and 10 are expressed at the cell surface andmainly recognize bacterial products. On the other hand,TLRs 3, 7, 8 and 9 are located almost exclusively inintracellular compartments and are specialized in recog-nition of nucleic acids, with self versus non-self discrim-ination provided by the exclusive localization of theligands rather than their different molecular structurefrom that of the host. TLRs 2 and 4 can be located bothon the cell surface and intracellular [2,3]. In this study,TLRs will be divided in two groups: TLRs 1, 2, 4, 5, 6and 10 as extracellular TLRs and TLRs 3, 7, 8 and 9 asintracellular TLRs and nucleic acid sensors.TLRs are conserved through evolution, from Drosophila
to mammals (reviewed at [4]), because of its essential rolein innate immunity. However, there are significant distinc-tions between intracellular and extracellular TLRs. Intra-cellular TLRs do not accept much variability, because theyhave evolved under strong purifying selection [5]. Virusescan only be detected through their nucleic acids; thereforeintracellular TLRs have an essential non-redundant role inhost survival. Moreover, mutations in those TLRs couldend up with an autoimmune disease against own nucleicacids or with high susceptibility to some viral infections.On the other hand, membrane or extracellular TLRshave evolved under less evolutionary pressure, due to theycan recognize one pathogen through different PAMPs(immunological redundancy). So they show a higher rateof damaging non-synonymous and STOP mutations.Although infective pressure that has reached these
molecules is one of the main mechanisms of evolution,it is not the only one. Non-adaptative evolution has alsoan important role, through genetic drift, bottlenecks andmigratory routes [6]. This kind of evolution should beseen in dogs, due to a first bottleneck with domestica-tion and a second one for the artificial selection ofbreeds [7]. For these reason it should be taken intoaccount the need for dealing with different breeds, and
even with the wolf, for the analysis of canine TLRpolymorphism.In humans, many studies are addressed to find out pos-
sible links between some TLR polymorphism and suscep-tibility or resistance to disease (for a review see [6]). Somegenetic variants in TLRs in dogs could be associated todifferences in breed innate immune response to pathogensand thus to susceptibility to infections or autoimmunediseases. So far, polymorphisms in TLR4 and TLR5 havebeen associated with Inflammatory Bowel disease (IBD) inGerman Shepherd dogs (GSD) [8], but only protectiveSNPs from TLR5 have been associated with IBD in other38 dog breeds [9] There is therefore great interest in thecharacterization of canine TLRs. TLR5 risk-associatedhaplotype for canine IBD confers hyper-responsiveness toflagellin [10]. Moreover, dogs with spontaneous IBD ex-hibit alterations in the enteric microbiota, which bearresemblance to dysbiosis reported in humans with chronicintestinal inflammation [11].Although no other polymorphisms have been associ-
ated to illness in dogs until date, some studies havereported differential expression of some TLRs related toinflammatory or infectious diseases, such as TLR2 inIBD [12], TLRs 2, 4, 5 and 9 in chronic enteropathies inGerman Shepherd [13,14]; TLR4 in osteoarthritis [15]and in infected canine endometrium [16]; TLRs 1-4, 6-10 in sino-nasal aspergillosis and idiopathic lymphoplas-macytic rhinitis [17]; and TLR2 and TLR9 in Leishmaniainfected dogs [18,19].So our aim is the analysis of genetic variation in exonic
regions of canine TLRs by massive sequencing, focusing innon-synonymous substitutions and their segregation indifferent dog breeds and wolf populations. A secondobjective is to design and validate a TaqMan OpenArray®plate of SNPs with a possible functional effect in the pro-tein (STOP, frameshift and non-synonymous codons).High-throughput genotyping of canine TLRs with thisTaqMan OpenArray® plate will allow studying the associ-ation of non-synonymous variants with individual differ-ences in immune response, their relationship with eitherthe commensal or the disease associated microbiota andTLR evolution in the canine genome.
ResultsWe have identified 156 new variants in canine TLRs bymassive sequencing after the enrichment of exonic re-gions. DNAs from 335 dogs (seven breeds) and 100 wolves(two populations) were pooled in 16 pools and sequencedin 2 lanes of Illumina Hiseq, with a mean coverage valueof 15,162.23. Dog breeds included were Beagle, Labrador,German Shepherd, Yorkshire, French Bulldog, Boxer andShar Pei. Wolves included were Iberian (Canis lupussignatus) and Russian (European grey wolf, Canis lupuslupus). A total of 204 variants were detected: 193 SNP and
Cuscó et al. Canine Genetics and Epidemiology 2014, 1:11 Page 2 of 12http://www.cgejournal.org/content/1/1/11
11 insertions or deletions (1 to 18 bases). Only one of theindels (insertion/deletion) mapped to an exonic region(TLR7 3′ UTR), meanwhile the others were mapping tointronic regions (5 out of 11) and intergenic regions up-stream or downstream a TLR gene (5 out of 11). The SNPsidentified were classified by functional annotation fromENSEMBL [20] (effect and effect impact): 73 synonymousvariants, 64 non-synonymous variants and 67 modifiervariants which include intergenic (upstream and down-stream a TLR gene), intronic and 3′ UTR (untranslated re-gion) variants (see Table 1). None of the variants detectedin the pools analyzed had a high effect (STOP codon,frameshift mutation or splicing) on the protein function.The ratio of SNP discovery was 76.5% (in relation toCanFam 3.1); 156 out of 204 variants identified were new:43/64 non-synonymous variants (nsSNP), 56/73 synonym-ous variants (synSNP) and 57/67 modifier variants.Genetic variation differs among all TLRs. Variants de-
tected in either extracellular or intracellular canine TLRsby massive sequencing and its classification accordingtheir effect in the protein are shown in Table 1. TLR5 genepresents the highest polymorphism, with 28 synonymouschanges and 23 non-synonymous changes (Additionalfile 1, Table 1), although it also codifies for the longestannotated protein (1422 aminoacids).Table 2 shows the aminoacid (AA) change ratio, which
are AA changes caused by nsSNPs or frameshift mutationsdivided by total number AA for each one of the TLRs.The AA change ratio confirms that indeed TLR5 andTLR4 are the most polymorphic ones. On the other hand,TLR3 seem to be the most conserved receptor, justpresenting one AA change in 905 AA.
Non-synonymous SNPsA more exhaustive analysis was performed for the 64nsSNP detected through massive sequencing, becausethey are expected to have a greater effect on the proteinfunction. First, a glimpse on allelic frequencies of the
nsSNP was performed. The frequencies of the alternativeallele for all 64 nsSNPs are shown for each breed andwolf pools in Additional file 2.Allelic frequencies for alternative variants in nsSNPs
differ among breeds. Beagle and Russian wolf are themost variable pools, with 35 out of 64 nsSNPs segregat-ing. Some of the variants identified are breed-specific (8out of 64) or wolf-specific (4 out of 64). Most of thebreed specific variants are found in TLR5 and TLR4,which as seen before, are the most polymorphic TLRs.German Shepherd dogs (GSD) and wolf share 3 nsSNPs,all located in TLR4. The same happens with Shar Peiand wolf, they share 3 nsSNPs located in TLR2, TLR5and TLR6.SNPs with a MAF (Minor allele frequency) <0.05 have
been considered to be fixed in the cohort [21]. Usually itis the reference allele the one which is fixed, but in somecases (perhaps due to bad annotation of the SNP) is thealternative one. Iberian wolves’ cohort is the one withmore fixed variants, with only 24 out of 64 that aresegregating, followed by Yorkshire and Boxer, with 25out of 64 segregating variants.
Predicted impact of canine TLRs amino acid substitutionsPolyphen-2 [22,23], SIFT [24,25], and PROVEAN [26,27]tools were used in order to predict the effect of eachnsSNP in the protein structure. Each of these tools uses adifferent algorithm to predict the consequence of the ami-noacid change on the protein and classifies it as benign/tolerated/neutral or damaging/affect protein function/deleterious (for more detail, see Methods). 28 out of 64nsSNPs were predicted to have an effect on the proteinstructure by at least one of the tools used (Table 3). Whenfrequency of the alternative variant was high for all thecohorts tested, the alternative allele was exchanged withthe reference allele in ENSEMBL sequences [20] in orderto perform the Polyphen-2 analysis with the less frequentvariant as the “alternative variant”. Therefore, SNPs with
Table 1 Variants detected in canine TLRs by massive sequencing
Extracellular TLRs Intracellular TLRs
Effect impact SNP effect TLR1 TLR2 TLR4 TLR5 TLR6 TLR10 TLR3 TLR7 TLR8 TLR9 Total
Low Syn coding 2 5 4 28 2 5 6 6 8 7 73
Moderate Non-syn coding 4 3 12 23 4 3 1 3 4 7 64
Total coding SNP (cSNP) 6 8 16 51 6 8 7 9 12 14 137
Modifier Downstream 1 4 0 0 2 3 0 0 2 1 12
Intron 0 0 6 3 0 0 10 7 0 2 28
Upstream 0 6 0 0 2 1 1 1 1 0 12
UTR 3′ 0 0 4 0 0 0 0 10 0 0 14
Total non coding SNP (ncSNP) 1 10 10 3 4 4 11 18 3 3 66
Total SNP 7 18 26 54 10 12 18 27 15 17 204
Variants are classified according to their effect on the protein and their spread along cell surface or intracellular TLRs.
Cuscó et al. Canine Genetics and Epidemiology 2014, 1:11 Page 3 of 12http://www.cgejournal.org/content/1/1/11
frequencies greater than 0.25 for the alternative allele weretested also for the annotated reference allele. Then, 3 moreSNPs were predicted to affect the protein structure (indi-cated as reference on the column dbSNP ID in Table 3).When considering also these ones, 31 out of 64 nsSNPs(48%) were predicted to have an impact on the proteinstructure. Results from Polyphen-2, SIFT and PROVEANwere convergent in predicting damaging effects for 8 outof 31 nsSNPs (27%). On the other hand, 6 out of 64nsSNPs were not correctly predicted, giving unknown orlow confidence results, because they were not aligning toenough similar sequences to give a reliable result. Curi-ously most of this nsSNPs were located on the N-terminalregion of TLR5.Protein structure of the canine TLRs was assessed
using SMART [28,29], which predicts domains takinginto account aminoacid sequences: 6 out of 31 nsSNPspredicted to be damaging in canine TLRs were found tobe in a Leucine Rich Repeat C-terminal (LRRCT) orreally close to it. Only 1 out of 31 was found to affectTIR domain in TLR 5, other 2 were found to be reallyclose to this domain in TLR5 and 10. With the excep-tion of these last ones, nsSNPs were in most caseslocated in the sensor domain of TLRs (Table 3).Frequencies in Table 3 are an average of all dog pools
tested and both wolf populations respectively, so all vari-ants are polymorphic (MAF > 0.05) at least in one breed.17 out of 31 show a MAF > 0.05 when considering theaverage frequencies in all the pools together (15 out of31 with MAF > 0.05 in wolf populations). However, asmentioned above, frequencies of nsSNPs differ amongbreeds (see Additional file 2). It’s worthy to note thedifferences on the alternative allele frequency observedfor the 8 nsSNPs that were predicted to affect proteinfunction by the three tools used (Figure 1).
TaqMan open array design and SNP validationA TaqMan OpenArray® plate has been developed for thevalidation of the nsSNPs by individual genotyping (Table 4).This panel contains (i) 27 out of 31 nsSNPs that were pre-dicted to have an impact on the protein structure (4 wolf-specific SNPs were not considered for the array: TLR1A525V, TLR5 N833K, TLR6 P579L and TLR10 F787L; seeTable 3); (ii) 28 out of the 33 remaining nsSNPs segregat-ing in dogs (5 SNPs that were not suitable for a correctprimer design were rejected for posterior analysis: TLR4T36A, TLR4 T36I, TLR5 T243A, TLR5 Q213R and TLR9A442V); and (iii) 4 frameshift and 4 non-synonymous TLRpolymorphisms described on CanFam 3.1 but not detectedin our cohorts (see Table 4). One of the non-synonymousvariants added (rs23572381, TLR1 N634K) was designedwith two different TaqMan assays due to the presence ofother variants close to the interrogated SNP.A total of 99 DNA samples of the first massive sequen-
cing pools were chosen to be individually genotyped inorder to validate the SNPs with the TaqMan Open Array®designed: 15 Beagle, 15 Boxer, 14 French bulldog, 15Labrador, 15 German Shepherd dog, 13 Yorkshire and 12Shar Pei were used. One Shar-Pei and 2 Yorkshires do notpass the quality control for samples (call rate > 0.9) andwere removed from the posterior analysis. Finally, analysiswas performed with a total of 96 individuals. Fifty-nineout of the 64 SNPs (92%) included in the OpenArray havebeen successfully validated and all of them had a call rategreater than 0.9.Some downstream analyses have been performed with
the individual genotypes. However, it should be takeninto account that these are just preliminary results,which need to be validated with larger cohorts of dogs.All the TLR SNPs were in Hardy-Weinberg Equilibrium
(HWE) on Beagle, Boxer, German Shepherd, Labrador and
Table 2 Total number of variants affecting protein in extracellular and intracellular TLRs
Canine gene ENSEMBL protein ID Protein length (aa) AA change ratioa
Extracellular TLRs
TLR1 ENSCAFP00000032660 790 1/113
TLR2 ENSCAFP00000012269 785 1/196
TLR4 ENSCAFP00000031395 833 1/69
TLR5 ENSCAFP00000016726 1422 1/53
TLR6 ENSCAFP00000023836 797 1/199
TLR10 ENSCAFP00000023840 807 1/269
Intracellular TLRs
TLR3 ENSCAFP00000011004 905 1/905
TLR7 ENSCAFP00000017193 1121 1/374
TLR8 ENSCAFP00000031505 1038 1/260
TLR9 ENSCAFP00000030804 1032 1/129
Variants from CanFam 3.1 have been added to variants identified by massive sequencing in this table. aAA change ratio: aminoacid changes caused by nsSNPs orframeshift mutations divided by the length of the protein in aminoacids.
Cuscó et al. Canine Genetics and Epidemiology 2014, 1:11 Page 4 of 12http://www.cgejournal.org/content/1/1/11
Shar-Pei. In Yorkshire, TLR10 has two SNPs in linkage dis-equilibrium which are not in HWE, one of them is predictedto affect protein function by the algorithms tested (Table 3).French Bulldog was the breed that had more SNPs that didnot follow HWE proportions, with 3 SNPs in TLR4 and oneSNP in TLR5 (Table 5). TLR7 and 8 were not includedbecause they are both located in chromosome X.
Principal components analysis (PCA) combined datafrom the individual genotypes obtained for the subset ofSNPs which were not in linkage disequilibrium. It was usedto illustrate if dogs cluster by breed for genetic variants inTLRs. The first two components from the PCA have beenplotted in Figure 2. Visual examination of this plotshows overlapping for most breeds, excluding Labrador
Table 3 Non-synonymous SNPs predicted to impact protein function either by Polyphen-2, SIFT or PROVEAN
Caninegene
Position SNP dbSNP ID AASubst
Proteindomaina
Polyphen-2result
SIFT result Proveanresult
Variant freq(dog)b
Variant freq(wolf)b
EXTRACELLULAR TLRs
TLR1 3:73542337 G/T rs23585044 S29I ncp Pos. damaging Tolerated Neutral 0,36 0,77
3:73543092 T/G new S281A ncp Pos. damaging Tolerated Neutral 0,06 0,00
3:73543825 C/T new A525V LRRCT2 Pos. damaging Tolerated Deleterious 0 0,11
TLR2 15:51463020 C/A rs22410121 S46Y ncp Pos. damaging Tolerated Neutral 0,10 0,00
15:51464430 C/T new S516L ncp Prob. damaging Aff. function Deleterious 0,14 0
TLR3 16:44623632 C/G new E176D ncp Pos. damaging Tolerated Neutral 0,16 0,12
TLR4 11:71356420 C/T reference1 A8V ncp Prob. damaging Tolerated Neutral 0,77 0,57
11:71360887 G/A new V82M ncp Pos. damaging Tolerated Neutral 0,09 0,15
11:71364581 T/C rs22145736 L167P ncp Prob. damaging Aff. function Deleterious 0,15 0
11:71364681 A/C reference1 Q200H ncp Pos. damaging Aff. function Neutral 0,88 0,23
11:71365810 A/G new T577A LRRCT3 Pos. damaging Aff. function Neutral 0,01 0
TLR5 38:23702837 C/T rs9070447 R269C ncp Prob. damaging Aff. function Neutral 0,19 0,01
38:23702918 G/A new V296I ncp Pos. damaging Tolerated Neutral 0,05 0,26
38:23703629 G/A new G533S ncp Prob. damaging Tolerated Neutral 0,02 0
38:23704331 G/T new D767Y ncp Prob. damaging Tolerated Deleterious 0,04 0
38:23704531 C/G new N833K LRRCT Pos. damaging Tolerated Deleterious 0 0,06
38:23704562 C/T new R844C LRRCT Pos. damaging Tolerated Neutral 0,04 0,39
38:23704581 C/T reference1 S850L LRRCT Prob. damaging Aff. function Deleterious 0,68 0,02
38:23704695 T/G new F888C lowcomplexity
Prob. damaging Aff. function Deleterious 0,04 0
38:23705081 C/T new H1017Y TIR Benign Aff. function Neutral 0,02 0
38:23705264 G/A new A1078T TIR4 Pos. damaging Aff. function Neutral 0,07 0,00
TLR6 3:73521250 A/G new Y182C ncp Prob. damaging Tolerated Deleterious 0,01 0,09
3:73522074 C/T new L457F ncp Prob. damaging Aff. function Deleterious 0,03 0
3:73522242 G/A rs23570247 D513N ncp Pos. damaging Tolerated Neutral 0,73 1,00
3:73522441 C/T new P579L LRRCT Pos. damaging Aff. function Deleterious 0,01 0,07
TLR10 3:73569402 C/T rs23518574 T361M ncp Prob. damaging Aff. function Deleterious 0,13 0,12
3:73570681 T/A new F787L TIR5 Pos. damaging Lowconfidence
Neutral 0,00 0,39
INTRACELLULAR TLRs
TLR8 X:9397240 T/C new V157A ncp Pos. damaging Aff. function Deleterious 0,06 0
TLR9 20:37544129 G/A new V87I ncp Benign Aff. function Neutral 0,02 0
20:37546230 C/T new P787L ncp Pos. damaging Tolerated Neutral 0,22 0,24
20:37546454 C/T new R862W ncp Prob. damaging Tolerated Neutral 0,2 0
In italics, SNPs that are predicted to have an effect on protein function by the three algorithms. ancp, no confident prediction. bObserved frequency by massivesequencing. 1reference allele tested as the alternative in the SNP. 2Leucine Rich Repeat C-terminal (LRRCT) domain predicted from aminoacid 528 to 582. 3LRRCTdomain predicted from aminoacid 579 to 629. 4TIR domain predicted from aminoacid 927 to 1074. 5TIR domain predicted from aminoacid 641 to 784.
Cuscó et al. Canine Genetics and Epidemiology 2014, 1:11 Page 5 of 12http://www.cgejournal.org/content/1/1/11
and perhaps German Shepherd, which seem to be moredifferentiated for these receptors.
DiscussionCanine breed specific variants in TLRs could be associated todifferences in breed innate immune response to pathogensand thus to susceptibility to infections or autoimmune dis-eases. So far, polymorphisms in TLR4 and TLR5 have beenassociated with IBD in German Shepherd dogs [8], but onlyprotective SNPs from TLR5 have been associated with IBDin other 38 dog breeds [9]. There is therefore great interest inthe characterization of canine TLRs. Different dog breeds and2 different populations of wolves (Iberian and Russian) wereincluded in the analysis to represent some of the majorphylogenetic radiations: Wolves, Ancient&Spitz breeds, Scenthounds, Working dogs, Mastiff-like dogs, Small Terriers andRetrievers [30]. A total of 204 variants have been discoveredand functionally annotated in exonic regions of canine TLRsby massive sequencing: 155 of the variants were new in rela-tion to the most recent annotation of the canine genome(CanFam 3.1; September 2012). Variants have been function-ally annotated and correspond to 64 non-synonymous vari-ants (43 new), 73 synonymous variants (56 new) and 67modifier variants (57 new). None of the variants detected inthe pools analyzed had a high effect (STOP codon, frameshiftmutation or splicing) on the protein function, although 4frameshift mutations are annotated in CanFam 3.1.SNPs functionally annotated as non-synonymous are
expected to have a greater effect on protein function, andtherefore a more exhaustive analysis was performed onthem. Although allelic frequencies for nsSNPs differamong breeds and 12.5% of them are breed-specific (6.25%are wolf specific), dogs from different breeds share mostnon-synonymous variants in TLRs.
A TaqMan OpenArray® plate containing 64 SNPs with apossible functional effect in the protein (4 frameshifts and60 nsSNPs) has been designed and validated. 55 out of 64SNPs contained in the OpenArray® plate have been identi-fied in this work through massive sequencing by HISEQ;the remaining 9 were obtained from CanFam 3.1.As shown in Figure 2, the individual genotypes are not
clustering by breed, with the exception of Labrador andGerman Shepherd dogs.The functional impact of non-synonymous variants in
dog TLRs was predicted using Polyphen-2, SIFT and PRO-VEAN. Knowing that TLRs are highly conserved recep-tors, it is not unexpected that half non-synonymousmutations in dogs have a benign effect, which agrees withresults from similar approaches in other non-primatespecies such as bovine [31]. In dogs, TLR5 is the one thatpresents more damaging non-synonymous mutations (pos-sibly damaging + probably damaging), followed by TLR4,both of them extracellular receptors.Results from SIFT and Polyphen-2 from some nsSNPs
located in TLR5 returned no output and no prediction(“unknown” or “low confidence”). In dogs, TLR5 was de-scribed as a longer protein compared to their homologs inother species. In CanFam 3.1 TLR5 has 1422 aminoacids,however other species like human, cow and pig have 858aa, 858 aa and 856 aa, respectively. A protein BLAST wasperformed with the extra 5′ and 3′ TLR5 fragments, butno result was obtained. Furthermore, the 5′ sequencebegins with ATG codon in the same phase as the initialcoding ATG in other species, whereas the 3′ sequenceeliminates the STOP codon due to some repeats intandem (data not shown). So, bad annotation of this genein CanFam3.1 is suggested. However, SNPs have beenfound in this region. In fact, there are 2 SNPs that hadalready been wrongly described as an aminoacid change
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
TLR2S516L
TLR4L167P
TLR5S850L
TLR5F888C
TLR6L457F
TLR6P579L
TLR8V157A
TLR10T361M
Alle
le f
req
uen
cyBeagle
Labrador
German S.
Yorkshire
French B.
Boxer
Shar Pei
Figure 1 Breed allelic frequencies for the 8 nsSNP with a damaging prediction from Polyphen-2, SIFT and PROVEAN.
Cuscó et al. Canine Genetics and Epidemiology 2014, 1:11 Page 6 of 12http://www.cgejournal.org/content/1/1/11
Table 4 Non-synonymous SNPs and frameshift mutations of canine TLRs in the TaqMan Open Array plate
Canine gene SNP Chr:bp position dbSNP ID AA Subst Previous detecteda Validated?
TLR1 G/T 3:73542337 rs23585044 S29I Massive seq YES
T/G 3:73543092 new S281A Massive seq YES
G/A 3:73543185 new V312I Massive seq YES
T/A 3:73544153 rs23572381 N634K1 CanFam 3.1 YES2
T/A 3:73544153 rs23572381 N634K1 CanFam 3.1 NO
G/A 3:73544221 rs23572380 S657N CanFam 3.1 YES2
TLR2 C/A 15:51463020 rs22410121 S46Y Massive seq YES
A/0 15:51464076 rs8958543 A398- CanFam 3.1 YES2
C/T 15:51464430 new S516L Massive seq YES
C/T 15:51464700 new T606M Massive seq YES
TLR3 C/G 16:44623632 new E176D Massive seq YES
TLR4 T/C 11:71356420 rs22120766 V8A Massive seq NO
G/C 11:71360743 rs22157966 A34P Massive seq YES
G/A 11:71360887 new V82M Massive seq YES
T/C 11:71364581 rs22145736 L167P Massive seq YES
C/A 11:71364681 rs22189454 H200Q Massive seq YES
A/G 11:71364769 rs22189456 K230E Massive seq YES
G/A 11:71365120 new A347T Massive seq YES
A/T 11:71365652 rs22124023 E524V Massive seq YES
A/G 11:71365810 new T577A Massive seq YES
G/A 11:71365888 rs22123995 E603K Massive seq YES
TLR5 G/A 38:23702193 rs24029590 G54E CanFam 3.1 NO
0/C 38:23702251 rs9070448 -74C CanFam 3.1 YES2
A/G 38:23702514 rs9070450 Y161C CanFam 3.1 NO
A/C 38:23702539 new E169D Massive seq YES
G/A 38:23702562 new S177N Massive seq YES
G/C 38:23702640 rs9070451 R203P Massive seq YES
T/C 38:23702684 rs9070452 W218R Massive seq NO
C/T 38:23702837 rs9070447 R269C Massive seq YES
G/A 38:23702918 new V296I Massive seq YES
T/C 38:23703180 new L383S Massive seq YES
G/A 38:23703237 new R402Q Massive seq YES
G/A 38:23703279 new R416Q Massive seq YES
T/0 38:23703591 rs9125247 T520- CanFam 3.1 YES2
G/A 38:23703629 new G533S Massive seq YES
G/A 38:23704233 new R734Q Massive seq YES
G/T 38:23704331 new D767Y Massive seq YES
C/T 38:23704562 new R844C Massive seq YES
T/C 38:23704581 rs24029975 L850S Massive seq YES
T/G 38:23704695 new F888C Massive seq YES
G/A 38:23704718 new A896T Massive seq YES
C/T 38:23705081 new H1017Y Massive seq YES
Cuscó et al. Canine Genetics and Epidemiology 2014, 1:11 Page 7 of 12http://www.cgejournal.org/content/1/1/11
(in ENSEMBL) moreover in this study 5 more SNPshave been detected. So it would be interesting to eitherdetermine the existence and functionality of these extrafragments in canine TLR5 cDNA or correctly annotate itin CanFam 3.1.Intracellular TLRs, which detect nucleic acids, have
less nsSNPs (15), moreover these are predicted to be lessdamaging variants than those identified in extracellularTLRs, suggesting that intracellular TLRs are selectivelyconstrained. TLR9 is the intracellular TLR that accepts
more nsSNPs in dogs, but the predicted effect of thesensSNPs is usually benign.These results agree with previously reported data reveal-
ing major differences in the intensity of selection actingupon the different members of the TLR family. DifferentTLRs differ in their immunological redundancy, reflectingtheir distinct contributions to host defense [5,32]. Intracel-lular TLRs act as nucleic acid sensors and have evolvedunder strong purifying selection, indicating their essentialnon-redundant role in host survival. Higher rates of
Table 4 Non-synonymous SNPs and frameshift mutations of canine TLRs in the TaqMan Open Array plate (Continued)
G/A 38:23705090 new G1020S Massive seq YES
G/A 38:23705178 new R1049Q Massive seq YES
G/A 38:23705264 new A1078T Massive seq YES
TLR6 A/G 3:73521250 new Y182C Massive seq YES
C/T 3:73522074 new L457F Massive seq YES
G/A 3:73522242 rs23570247 D513N Massive seq YES
TLR7 C/G X:9334108 new A16G Massive seq YES2
C/A X:9355727 new F167L Massive seq YES
C/T X:9358423 new P1066L Massive seq YES
TLR8 T/C X:9397240 new V157A Massive seq YES
G/A X:9397663 new R298Q Massive seq YES
G/A X:9398094 rs24607342 G442S Massive seq YES
G/A X:9398827 rs24607358 R686H Massive seq YES
TLR9 G/A 20:37544129 new V87I Massive seq YES
0/A 20:37544851 rs9188882 -328A CanFam 3.1 YES2
A/G 20:37545011 new K381E Massive seq YES
C/A 20:37545245 new P459T Massive seq YES
A/G 20:37546031 rs22882109 S721G Massive seq YES
C/T 20:37546230 new P787L Massive seq ND3
C/T 20:37546454 new R862W Massive seq YES
TLR10 C/T 3:73569402 rs23518574 T361M Massive seq YES
A/G 3:73570094 new M592V Massive seq YESaMassive seq indicates a SNP variant detected in our cohorts. An “rs” name is indicated in dbSNP ID if the SNP is annotated in CanFam 3.1. 1SNP considered twicewith a different surrender SNP in order to detect it. 2Assay has been validated technically, although not genetically because all individuals have only the referenceallele. 3ND (not determined), there are incongruent results: massive sequencing showed that this SNP was present at a frequency of 0.2 in all breeds tested,whereas it has not been genotyped through TaqMan OA plate.
Table 5 SNPs in different breeds that are not in Hardy-Weinberg Equilibrium (p < 0.05)
Breed Canine gene AA change SNP Genotypesa p-value SNP predictionb
Yorkshire TLR10 T361M C/T 1/1/9 0.0416978 Prob. damaging
Yorkshire TLR10 M592V A/G 1/1/9 0.0416978 Benign
French B. TLR4 V82M G/A 4/2/8 0.0099493 Pos. damaging
French B. TLR4 H200Q C/A 6/1/7 0.0013535 Pos. damaging*
French B. TLR4 K230E A/G 6/1/7 0.0013535 Benign
French B. TLR5 S177N G/A 0/11/3 0.0154748 Benignagenotypes, indicate genotype count for reference homozygotes, heterozygotes and alternative homozygotes. bSNP prediction, using Polyphen-2 classification.*possibly damaging when reference allele is tested as alternative in the SNP.
Cuscó et al. Canine Genetics and Epidemiology 2014, 1:11 Page 8 of 12http://www.cgejournal.org/content/1/1/11
damaging non-synonymous and nonsense mutations aretolerated in cell-surface or extracellular TLRs, whichrecognize compounds other than nucleic acids, suggestinga higher redundancy.Location of the SNPs in the protein was approached
using the software SMART [28], which identifies TLRdomains using the aminoacid sequence. The intracellularTIR domain is highly conserved between different TLRsand species due to its involvement in intracellular signal-ing [33]. Also in dogs, TIR domains have few SNPs; onlyone is present in the predicted TIR domain (TLR5H1017Y) and another two are really close to it (TLR5A1078T and TLR10 F787L). Extracellular domains ofTLRs are those that recognize PAMPs, and they have anenhanced susceptibility to mutate adapting to differentmicrobiologic environments [33]. It can also be seen thata high number of mutations (some with damaging effects)are located in LRR domains, which form the extracellulardomain of TLRs.So far, polymorphisms in TLRs have been associated with
Inflammatory Bowel disease (IBD) in German Shepherddogs (GSD) and in other breeds. Variants in TLR5 previ-ously reported to be associated to IBD (G22A, C100T andT1844C from [8,9]) have been also detected in our cohortsand correspond with TLR5 T243A, TLR5 R269C and TLR5L850S respectively.SNP G22A, where the risk allele is A in G22A (corre-
sponding to Thr in TLR5 T243A as named in this work),is found to be an additive allele. So when a GSD ishomozygous for the risk allele it has more susceptibilityto suffer IBD. This risk allele is not segregating in ourGSD cohort. This could be due to the difference in thegeographical origin of the GSD cohort between bothstudies. In [9] GSD are from UK, whilst our cohort isfrom Spain. SNPs C100T and T1844C were found tobe significantly protective against canine IBD in many
breeds [9]. The frequencies of the protective alleles (T inC100T and T in T1844C or Cys in TLR5 R269C and Leuin TLR5 L850S as named in this work) differ amongbreeds (Figure 3), with a frequency higher than 0.5 inYorkshire and GSD.
ConclusionsPolymorphisms in the exonic regions of canine TLRshave been characterized by massive sequencing and156 out of 204 variants identified were new: 43/64non-synonymous variants, 56/73 synonymous variantsand 57/67 modifier variants. None of the variantsdetected in the pools analyzed had a high effect (STOPcodon, frameshift mutation or splicing) on the proteinfunction.A TaqMan OpenArray® plate containing 64 SNPs with
a possible functional effect in the protein (4 frameshiftsand 60 nsSNPs) has been designed and validated toallow the high throughput genotyping of canine TLRs.
MethodsEthics statementThe dogs in the study were examined during routinaryveterinary procedures by the veterinary clinics participat-ing in the study. All samples were collected for routinediagnostic and clinical purposes. The samples wereobtained during veterinary procedures that would havebeen carried out anyway and DNA was extracted fromresidual surplus of samples and used in the study withverbal owner consent. This is a very special situation inveterinary medicine. As the data are from client-owneddogs that underwent normal veterinary exams, there wasno “animal experiment” according to the legal defini-tions in Spain and the United Kingdom, and approval byan ethical committee was not necessary.
-0.35
-0.30
-0.25
-0.20
-0.15
-0.10
-0.05
0.00
0.05
0.10
0.15
-0.20 -0.10 0.00 0.10 0.20 0.30 0.40
PC
1
PC 2
Beagle
Boxer
French B.
German S.
Labrador
Shar Pei
Yorkshire
Figure 2 Principal Component Analysis (PCA) plot of the two first components for canine TLRs.
Cuscó et al. Canine Genetics and Epidemiology 2014, 1:11 Page 9 of 12http://www.cgejournal.org/content/1/1/11
DNA sourcesSamples available from the DNA bank at the SVGM(Molecular Genetics Veterinary Service, UAB) were used.Total DNA from blood cells had been extracted either asdescribed elsewhere [34] or using QIAamp DNA Mini Kit(Qiagen).DNAs from 7 different dog breeds, including 50 Beagle,
50 German Shepherd, 50 Yorkshire, 35 French bulldog, 75Boxer, 50 Labrador and 25 Shar Pei were used. All thedogs included in this study are from Spain region, andcome from hospital population or normal pet dogs. Also 2different populations of wolves, with 50 Iberian (Canislupus signatus) and 50 Russian (European grey wolves,Canis lupus lupus), were analyzed. DNA pools wereprepared with 200 ng of DNA from 25 unrelated dogs(with the exception of one pool of French bulldogs, withonly 10 dogs). Two pools of each breed were analyzed, inexception of Shar Pei (only 1 pool) and Boxer (3 pools).Pools of wolves were of 50 individuals.Some DNA samples of the first massive sequencing
analyses were chosen to be individually genotyped in orderto validate SNPs in the TaqMan Open Array® designed(15 Beagle, 15 Boxer, 14 French bulldog, 15 Labrador, 15German Shepherd, 13 Yorkshire and 12 Shar Pei).
Exon capture and massive sequencing for SNP discoveryTwenty exonic regions of 10 canine TLR genes annotatedin CanFam 2.0 were chosen to perform the enrichment(see Additional file 3 with corresponding coordinates inCanFam 3.1).Oligonucleotides were first automatically designed for
the enrichment of selected regions [35]. Regions rejectedin the automated design, because of the presence of gaps,repeats or shorter sizes than required (at least 120 nucleo-tides) were manually redesigned. Finally 235 ultra-long120-mer biotinylated cRNA baits were designed to capturethe exonic regions of canine TLRs (28,200 bases) by theAgilent Sure Select technique. High-throughput sequencing
was performed using 2 lanes of Illumina HISEQ, with8-labelled pools each, at CNAG (Centre Nacional d’AnàlisiGenòmica, Barcelona, Spain).Sequences obtained were mapped to CanFam 3.1 (re-
leased September 2012). All pools were analysed togetherfor variant calling, for better comparison. Alternativevariant frequencies were estimated for each breed pooland wolf populations. The variants were annotated withstatistical information from the Genome Analysis Tool Kit(GATK) and functional annotations were added fromEnsembl using snpEff [36].
Prediction of functional impact of non-synonymous SNPsThe functional impact of non-synonymous mutationsdetected was predicted using Polyphen-2 [22,23], SIFT[25,24] and PROVEAN [27,26]. When the mean frequencyof an alternative variant on the dog population analyzedwas more than 0.25, both alleles of those SNPs weretested with algorithms mentioned before as referenceand alternative.Polyphen-2 classifies mutations in three categories: be-
nign, possibly damaging and probably damaging. Polyphenalgorithm considers protein structure and/or sequence con-servation information for each gene [23]. SIFT is based onthe evolutionary conservation of the amino acids withinprotein families performing multiple sequencing analysesusing PSI-Blast algorithm. Highly conserved positions tendto be intolerant to substitution, whereas those with a lowdegree of conservation tolerate most substitutions. There-fore, it classifies each non-synonymous polymorphism astolerated or affect protein function and provides also aconfidence measure [24]. PROVEAN introduced a region-based “delta alignment score” which measures the impactof an amino acid variation not only based on the aminoacid residue at the position of interest but also the qualityof sequence alignment derived from the neighborhoodflanking sequences. It classifies variants either as neutral ordeleterious [26].
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
G22A C100T T1844C
Beagle
Labrador
German S.
Yorkshire
French B.
Boxer
Shar Pei
Wolf
Risk allele in GSD Protective alleles (T) in GSD and other breeds
Figure 3 Observed allele frequency of the alleles related with IBD in our pools (A in G22A, and T in both C100T and T1844C).
Cuscó et al. Canine Genetics and Epidemiology 2014, 1:11 Page 10 of 12http://www.cgejournal.org/content/1/1/11
SMART was used in order to identify protein domainsof each TLR using their aminoacid sequence [28].
TaqMan OpenArray® designA TaqMan OpenArray® was designed for genotyping andvalidating 64 SNPs with a possible functional effect in theprotein. Selected SNPs and their surrounding sequences, 60nucleotides upstream and 60 nucleotides downstream wereintroduced in Custom TaqMan® Assay Design Tool website [37] from Life Technologies® to validate if the sequenceswere suitable for TaqMan assay design. Other SNPs in thecontext sequences were indicated with an “N” before theassays design. SNPs included are listed in Table 4.Analysis was performed with the TaqMan Genotyper
software v.1.3 (Applied Biosystems). Further analysis ofindividual genotypes was performed with SVS (version 7)of Golden Helix Inc. SNPs or samples that do not pass callrate >0.9 were removed for posterior analysis.
Additional files
Additional file 1: Total number of detected synonymous andnon-synonymous SNPs for each canine Toll-like receptor.
Additional file 2: Frequencies of coding variants obtained by massivesequencing. Frequency per breed (50 individuals approximately) and meanfrequency per dog and wolf species. Frequency of the SNP is represented asthe frequency of the alternative allele.
Additional file 3: Coordinates of exonic regions of 10 canine TLRgenes as annotated in CanFam 3.1.
Competing interestsA patent application has been filled in related to the use of some of themarkers described in the manuscript.
Authors’ contributionsAS, LF and OF designed the experiment. AS, LA, LF and OF supervised theproject and gave conceptual advice. LA collected the samples. AC carriedout the molecular genetic studies and designed the chip. AC and OFanalyzed the data and drafted the manuscript. AS, LA and LF edited themanuscript. All authors read and approved the final manuscript.
AcknowledgementsWe acknowledge Lorena Serrano (Vetgenomics) for helping with the collectionof the samples; Sophia Derdak and Sergi Beltran (CNAG) for the raw dataprocessing of the massive sequencing results; and Anna Mercadé (SVGM) fortechnical advice and support on the OpenArray design and validation.
Author details1Molecular Genetics Veterinary Service. Veterinary School, UniversitatAutònoma de Barcelona, Barcelona, Spain. 2Vetgenomics. Ed Eureka. Parc deRecerca UAB, Barcelona, Spain. 3Department of Clinical Sciences, CummingsSchool of Veterinary Medicine, Tufts University, North Grafton, MA, USA.
Received: 6 June 2014 Accepted: 23 September 2014Published: 22 October 2014
References1. Werling D, Jungi TW: TOLL-like receptors linking innate and adaptive
immune response. Vet Immunol Immunopathol 2003, 91:1–12.2. Kawai T, Akira S: The role of pattern-recognition receptors in innate
immunity: update on Toll-like receptors. Nat Immunol 2010, 11:373–384.3. Mogensen TH: Pathogen recognition and inflammatory signaling in
innate immune defenses. Clin Microbiol Rev 2009, 22:240–273.
4. Aderem A, Ulevitch RJ: Toll-like receptors in the induction of the innateimmune response. Nature 2000, 406:782–787.
5. Barreiro LB, Ben-Ali M, Quach H, Laval G, Patin E, Pickrell JK, Bouchier C,Tichit M, Neyrolles O, Gicquel B, Kidd JR, Kidd KK, Alcaïs A, Ragimbeau J,Pellegrini S, Abel L, Casanova J-L, Quintana-Murci L: Evolutionary dynamicsof human Toll-like receptors and their different contributions to hostdefense. PLoS Genet 2009, 5:e1000562.
6. Netea MG, Wijmenga C, O’Neill LAJ: Genetic variation in Toll-like receptorsand disease susceptibility. Nat Immunol 2012, 13:535–542.
7. Lindblad-Toh K, Wade CM, Mikkelsen TS, Karlsson EK, Jaffe DB, Kamal M,Clamp M, Chang JL, Kulbokas EJ, Zody MC, Mauceli E, Xie X, Breen M,Wayne RK, Ostrander EA, Ponting CP, Galibert F, Smith DR, DeJong PJ,Kirkness E, Alvarez P, Biagi T, Brockman W, Butler J, Chin C-W, Cook A, Cuff J,Daly MJ, DeCaprio D, Gnerre S, et al: Genome sequence, comparative analysisand haplotype structure of the domestic dog. Nature 2005, 438:803–819.
8. Kathrani A, House A, Catchpole B, Murphy A, German A, Werling D,Allenspach K: Polymorphisms in the TLR4 and TLR5 gene are significantlyassociated with inflammatory bowel disease in German shepherd dogs.PLoS One 2010, 5:e15740.
9. Kathrani A, House A, Catchpole B, Murphy A, Werling D, Allenspach K:Breed-independent toll-like receptor 5 polymorphisms showassociation with canine inflammatory bowel disease. Tissue Antigens2011, 78:94–101.
10. Kathrani A, Holder A, Catchpole B, Alvarez L, Simpson K, Werling D,Allenspach K: TLR5 risk-associated haplotype for canine inflammatorybowel disease confers hyper-responsiveness to flagellin. PLoS One 2012,7:e30117.
11. Suchodolski JS, Dowd SE, Wilke V, Steiner JM, Jergens AE: 16S rRNA genepyrosequencing reveals bacterial dysbiosis in the duodenum of dogswith idiopathic inflammatory bowel disease. PLoS One 2012, 7:e39333.
12. McMahon LA, House AK, Catchpole B, Elson-Riggins J, Riddle A, Smith K,Werling D, Burgener IA, Allenspach K: Expression of Toll-like receptor 2 induodenal biopsies from dogs with inflammatory bowel disease isassociated with severity of disease. Vet Immunol Immunopathol 2010,135:158–163.
13. Burgener IA, König A, Allenspach K, Sauter SN, Boisclair J, Doherr MG, JungiTW: Upregulation of toll-like receptors in chronic enteropathies in dogs.J Vet Intern Med 2008, 22:553–560.
14. Allenspach K, House A, Smith K, McNeill FM, Hendricks A, Elson-Riggins J,Riddle A, Steiner JM, Werling D, Garden OA, Catchpole B, Suchodolski JS:Evaluation of mucosal bacteria and histopathology, clinical diseaseactivity and expression of Toll-like receptors in German shepherd dogswith chronic enteropathies. Vet Microbiol 2010, 146:326–335.
15. Kuroki K, Stoker AM, Sims HJ, Cook JL: Expression of Toll-like receptors 2and 4 in stifle joint synovial tissues of dogs with or without osteoarthritis.Am J Vet Res 2010, 71:750–754.
16. Chotimanukul S, Sirivaidyapong S: Differential expression of Toll-likereceptor 4 (TLR4) in healthy and infected canine endometrium.Theriogenology 2011, 76:1152–1161.
17. Mercier E, Peters IR, Day MJ, Clercx C, Peeters D: Toll- and NOD-likereceptor mRNA expression in canine sino-nasal aspergillosis andidiopathic lymphoplasmacytic rhinitis. Vet Immunol Immunopathol 2012,145:618–624.
18. Figueiredo MM, Amorim IFG, Pinto AJW, Barbosa VS, Pinheiro LDJ, Deoti B,Faria AMC, Tafuri WL: Expression of Toll-like receptors 2 and 9 in cells ofdog jejunum and colon naturally infected with Leishmania infantum.BMC Immunol 2013, 14:22.
19. Melo LM, Perosso J, Almeida BFM, Silva KLO, Somenzari MA, de Lima VMF:Effects of P-MAPA immunomodulator on Toll-like receptor 2, ROS, nitricoxide, MAPKp38 and IKK in PBMC and macrophages from dogs withvisceral leishmaniasis. Int Immunopharmacol 2014, 18:373–378.
20. Ensembl genome browser 75: Canis lupus familiaris.[http://www.ensembl.org/Canis_familiaris/].
21. Nelson MR, Marnellos G, Kammerer S, Hoyal CR, Shi MM, Cantor CR, BraunA: Large-scale validation of single nucleotide polymorphisms in generegions. Genome Res 2004, 14:1664–1668.
22. PolyPhen-2: prediction of functional effects of human nsSNPs.[http://genetics.bwh.harvard.edu/pph2/index.shtml]
23. Adzhubei IA, Schmidt S, Peshkin L, Ramensky VE, Gerasimova A, Bork P,Kondrashov AS, Sunyaev SR: A method and server for predictingdamaging missense mutations. Nat Methods 2010, 7:248–249.
Cuscó et al. Canine Genetics and Epidemiology 2014, 1:11 Page 11 of 12http://www.cgejournal.org/content/1/1/11
24. Kumar P, Henikoff S, Ng PC: Predicting the effects of coding non-synonymousvariants on protein function using the SIFT algorithm. Nat Protoc 2009,4:1073–1081.
25. SIFT. [http://sift.jcvi.org/]26. Choi Y, Sims GE, Murphy S, Miller JR, Chan AP: Predicting the functional
effect of amino acid substitutions and indels. PLoS One 2012, 7:e46688.27. PROVEAN. [http://provean.jcvi.org/index.php]28. Letunic I, Doerks T, Bork P: SMART 7: recent updates to the protein domain
annotation resource. Nucleic Acids Res 2012, 40(Database issue):D302–D305.29. SMART. [http://smart.embl-heidelberg.de/]30. Vonholdt BM, Pollinger JP, Lohmueller KE, Han E, Parker HG, Quignon P,
Degenhardt JD, Boyko AR, Earl DA, Auton A, Reynolds A, Bryc K, Brisbin A,Knowles JC, Mosher DS, Spady TC, Elkahloun A, Geffen E, Pilot M,Jedrzejewski W, Greco C, Randi E, Bannasch D, Wilton A, Shearman J,Musiani M, Cargill M, Jones PG, Qian Z, Huang W, et al: Genome-wide SNPand haplotype analyses reveal a rich history underlying dogdomestication. Nature 2010, 464:898–902.
31. Fisher CA, Bhattarai EK, Osterstock JB, Dowd SE, Seabury PM, Vikram M,Whitlock RH, Schukken YH, Schnabel RD, Taylor JF, Womack JE, Seabury CM:Evolution of the bovine TLR gene family and member associations withMycobacterium avium subspecies paratuberculosis infection. PLoS One2011, 6:e27744.
32. Wlasiuk G, Nachman MW: Adaptation and constraint at Toll-like receptorsin primates. Mol Biol Evol 2010, 27:2172–2186.
33. Werling D, Jann OC, Offord V, Glass EJ, Coffey TJ: Variation matters: TLRstructure and species-specific pathogen recognition. Trends Immunol2009, 30:124–130.
34. Francino O, Altet L, Sánchez-Robert E, Rodriguez A, Solano-Gallego L,Alberola J, Ferrer L, Sánchez A, Roura X: Advantages of real-time PCR assayfor diagnosis and monitoring of canine leishmaniosis. Vet Parasitol 2006,137:214–221.
35. SureDesign. [https://earray.chem.agilent.com/suredesign/]36. Cingolani P, Platts A, Wang LL, Coon M, Nguyen T, Wang L, Land SJ, Ruden
DM, Lu X: A program for annotating and predicting the effects of singlenucleotide polymorphisms, SnpEff: SNPs in the genome of Drosophilamelanogaster strain w1118; iso-2; iso-3. Fly 2012, 6:80–92.
37. Custom TaqMan® Assay Design Tool - Life technologies.[https://www5.appliedbiosystems.com/tools/cadt/]
doi:10.1186/2052-6687-1-11Cite this article as: Cuscó et al.: Non-synonymous genetic variation inexonic regions of canine Toll-like receptors. Canine Genetics and Epidemiology2014 1:11.
Submit your next manuscript to BioMed Centraland take full advantage of:
• Convenient online submission
• Thorough peer review
• No space constraints or color figure charges
• Immediate publication on acceptance
• Inclusion in PubMed, CAS, Scopus and Google Scholar
• Research which is freely available for redistribution
Submit your manuscript at www.biomedcentral.com/submit
Cuscó et al. Canine Genetics and Epidemiology 2014, 1:11 Page 12 of 12http://www.cgejournal.org/content/1/1/11
68
3.2. Individual Signatures Define Canine Skin
Microbiota Composition and Variability
This chapter consists of the article entitled “Individual Signatures Define Canine Skin
Microbiota Composition and Variability” published in Frontiers in Veterinary Science in
February 2017 (4:6).
See Section 3.2.1 Annex 1: Erratum: Figure 3 to see the newest version of Figure 3,
The Supplementary Material for this article can be found online at:
http://journal.frontiersin.org/article/10.3389/fvets.2017.00006/full#supplementary-
material.
February 2017 | Volume 4 | Article 61
Original researchpublished: 06 February 2017
doi: 10.3389/fvets.2017.00006
Frontiers in Veterinary Science | www.frontiersin.org
Edited by: Carl James Yeoman,
Montana State University, USA
Reviewed by: Eric Altermann,
AgResearch, New Zealand Suleyman Yildirim,
Istanbul Medipol University International School of Medicine,
Turkey Jan Slapeta,
University of Sydney, Australia
*Correspondence:Anna Cuscó
anna.cusco@vetgenomics.com
Specialty section: This article was submitted to Veterinary Experimental and
Diagnostic Pathology, a section of the journal
Frontiers in Veterinary Science
Received: 17 October 2016Accepted: 17 January 2017
Published: 06 February 2017
Citation: Cuscó A, Sánchez A, Altet L, Ferrer L
and Francino O (2017) Individual Signatures Define Canine Skin
Microbiota Composition and Variability.
Front. Vet. Sci. 4:6. doi: 10.3389/fvets.2017.00006
individual signatures Define canine skin Microbiota composition and VariabilityAnna Cuscó1,2*, Armand Sánchez1, Laura Altet2, Lluís Ferrer3 and Olga Francino1
1 Molecular Genetics Veterinary Service (SVGM), Veterinary School, Universitat Autònoma de Barcelona, Barcelona, Spain, 2 Vetgenomics, Ed Eureka, Parc de Recerca UAB, Barcelona, Spain, 3 Department of Clinical Sciences, Cummings School of Veterinary Medicine, Tufts University, North Grafton, MA, USA
Dogs present almost all their skin sites covered by hair, but canine skin disorders are more common in certain skin sites and breeds. The goal of our study is to characterize the composition and variability of the skin microbiota in healthy dogs and to evaluate the effect of the breed, the skin site, and the individual. We have analyzed eight skin sites of nine healthy dogs from three different breeds by massive sequencing of 16S rRNA gene V1–V2 hypervariable regions. The main phyla inhabiting the skin microbiota in healthy dogs are Proteobacteria, Firmicutes, Fusobacteria, Actinobacteria, and Bacteroidetes. Our results suggest that skin microbiota composition pattern is individual specific, with some dogs presenting an even representation of the main phyla and other dogs with only a major phylum. The individual is the main force driving skin microbiota composition and diversity rather than the skin site or the breed. The individual is explaining 45% of the distances among samples, whereas skin site explains 19% and breed 9%. Moreover, analysis of similarities suggests a strong dissimilarity among individuals (R = 0.79, P = 0.001) that is mainly explained by low-abundant species in each dog. Skin site also plays a role: inner pinna presents the highest diversity value, whereas perianal region presents the lowest one and the most differentiated microbiota composition.
Keywords: skin, microbiota, microbiome, dog, canine, coat, skin site, 16s
inTrODUcTiOn
The skin is the living interface between an individual and the exogenous environment. It is covered with millions of microorganisms (1) interacting together with hosts’ cells and immune receptors to maintain the equilibrium (2). Bacteria are the most abundant microorganisms living on skin surface (3), and their whole population is defined as the microbiota. The high variability of the microbiota in the healthy skin has been captured during the last years using next-generation sequencing tech-niques [for a review, see Ref. (4)]. Marker-based approaches, mainly using 16S rRNA gene, focus on detecting who is living there—bacterial composition and diversity.
Main phyla inhabiting human skin are Actinobacteria, Firmicutes, Bacteroidetes, and Proteobacteria. A feature of human cutaneous microbiota is the existence of different microhabitats, which are characterized by the predominance of a specific taxa: sebaceous sites (occiput, glabella, alar crease, and manubrium) with Propionibacterium spp; moist sites (nare, axilla, and inguinal crease) with Staphylococcus and Corynebacterium spp; and dry sites (palms and butlock) with gram-negative microorganisms (5). According to the first extensive study reported, dogs harbor mainly the same
FigUre 1 | skin sites sampled per dog.
2
Cuscó et al. Skin Microbiota in Healthy Dogs
Frontiers in Veterinary Science | www.frontiersin.org February 2017 | Volume 4 | Article 6
phyla as human skin (6). Fusobacteria was also detected as a main phylum, when only considering the paws and the forehead (7) and also in a recent study considering the groins (8). In humans, the variation is higher among different microhabitat skin sites of the same individual than among skin sites from the same micro-habitat in different individuals (5, 9). Several differences among skin sites have been described in dogs (6), but to our knowledge, no microhabitats have been defined.
Different factors such as the environment, host genetic variation, lifestyle, or hygiene cause shifts on the microbial com-munities of the skin (10). These shifts on the microbiota structure and composition could establish a dysbiotic state, which if not recovered could result on a dermatologic affliction. Dysbiosis of the skin microbiota has been associated with several skin afflictions in humans, such as atopic dermatitis (11, 12), psoriasis (13, 14), and acne vulgaris (15). In canine microbiota studies, association between atopic dermatitis and microbiota has been assessed showing less richness on affected animals, either when considering bacteria (6, 16) or fungal communities (17). However, in allergen-induced canine atopic dermatitis, no significant differences on diversity were reported (8). Moreover, recent studies have reported significant increases of Staphylococcus and Corynebacterium in dogs with this disease (8, 16). Nevertheless, a better characterization of the cutaneous microbiota of healthy dogs seems to be necessary before understanding its role in disease conditions.
There is much less knowledge about the potential functions of the mammals’ microbiota. The potential function of a bacte-rial community can be assessed either directly, using shotgun metagenomics, or indirectly, using 16S data and a predictive software such as Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt) (18). Langille et al. used this tool with the Human Microbiome Project dataset (19) obtaining sufficiently accurate results, even for skin samples (18). In canine intestinal microbiota studies, shotgun metagen-omics has been used to study microbiota variability when feeding animals with two different diets (20), and PICRUSt was used in dogs suffering idiopathic inflammatory bowel disease (21). To our knowledge, no studies have assessed potential functions of the microbiota at the skin level.
Our aim was to characterize the composition and variability of the skin microbiota on healthy dogs, considering the breed—spe-cially the hair coat—the skin site, and the individual. We sampled nine healthy dogs from three breeds representing the diversity of canine hair coats: French Bulldog (FB; short hair), German Shepherd (GS; long hair with undercoat), and West Highland White Terriers (WHs; wired hair) (22). These three breeds were also selected because they are among the most predisposed to suffer from atopic dermatitis (23). We also aimed to predict the functional profile of the microbiota of different skin sites using PICRUSt.
MaTerials anD MeThODs
ethics statementThe dogs in the study were examined during routine veterinary procedures by the veterinary clinics participating in the study.
All samples were collected and used in the study with verbal owner consent. As the data are from client-owned dogs that underwent normal preventative veterinary examinations, there was no “animal experiment” according to the legal definitions in Spain, and approval by an ethical committee was not necessary.
individuals included and sample collectionA cross-sectional study was performed in nine healthy dogs to analyze skin microbiota variability in several skin sites, con-sidering the breed, the hair coat, and the individual. They were all pure-breed dogs ranging from 3 months to 12 years of age and from different households visiting the veterinary clinic for routine procedures (Table S1 in Supplementary Material). All of them lived in urban or periurban environment. Samples from three FBs (FB1, FB2, and FB3), three GSs (GS1, GS2, and GS3), and three West Highland WHs (WH1, WH2, and WH3) were included. Skin samples were collected from eight skin regions: chin, inner pinna, nasal skin, axilla, back, abdomen, interdigital area, and perianal region. These regions are named as 1, 2, 3, 4, 5, 6, 7, and 8, respectively (Figure 1). Samples were obtained by firmly rubbing each area using Sterile Catch-All™ Sample Collection Swabs (Epicentre Biotechnologies) soaked in sterile SCF-1 solution (50 mM Tris buffer (pH = 8), 1 mM EDTA, and 0.5% Tween-20). To minimize sample cross-contamination, the person sampling wore a fresh pair of sterile gloves for each indi-vidual. Swabs were stored at 4°C until DNA extraction, within the following 24 h.
Dna extractionBacterial DNA was extracted from the swabs using the PowerSoil™ DNA isolation kit (MO BIO) under manufacturer’s conditions, with one modification. At the first lysis step, the swab tip with the sponge was cut and introduced in the beads’ tube, until the first transference of the supernatant to a new tube. The remaining steps were performed as described by the manufacturer. DNA samples (100 µl) were stored at −20°C until further processing.
To assess for contaminations from the laboratory or reagents, a sterile swab tip was processed in the same conditions as the skin microbiota samples, giving negative results.
3
Cuscó et al. Skin Microbiota in Healthy Dogs
Frontiers in Veterinary Science | www.frontiersin.org February 2017 | Volume 4 | Article 6
Pcr amplification and Massive sequencingV1–V2 regions of 16S rRNA gene were amplified using the widely used primer pair F27 (5′-AGAGTTTGATCCTGGCTCAG-3′) and R338 (5′-TGCTGCCTCCCGTAGGAGT-3′). PCR mixture (50 uL) contained 5 µl of DNA template (~5 ng), 5 µl of 10× AccuPrime™ PCR Buffer II, 0.2 µM of each primer, and 1 U of AccuPrime™ Taq DNA Polymerase High Fidelity (Life Technologies). The PCR thermal profile consisted of an initial denaturation for 2 min at 94°C, followed by 30 cycles for 1 min at 94°C, 1 min at 55°C, 1 min at 72°C, and a final step for 7 min at 72°C. To assess possible reagent contamination, each PCR reaction included a no template control sample, which did not amplify. For each amplicon, quality and quantity were assessed using Agilent Bioanalyzer 2100 and Qubit™ fluorometer. Both primers included sequencing adaptors at the 5′ end, and forward primers were tagged with different barcodes to pool samples in the same sequencing reaction. Each pool contained 8 barcoded samples. A total of 9 pools were sequenced on an Ion Torrent Personal Genome Machine (PGM) with the Ion 318 Chip Kit v2 and the Ion PGM™ Sequencing 400 Kit (Life Technologies) under manufacturer’s conditions. The raw sequences have been deposited in NCBI under the Bioproject accession number PRJNA357691.
Quality control of the sequences and Operational Taxonomic Unit (OTU) PickingRaw sequencing reads were demultiplexed, quality filtered, and analyzed using QIIME 1.9.1 (24). Reads included had a length greater than 300 bp; a mean quality score above 25 in sliding window of 50 nucleotides; no mismatches on the primer; and default values for other quality parameters. Quality-filtered reads were clustered into OTUs at 97% similarity, using UCLUST (25) in an open reference approach for taxonomy analyses and a closed reference approach for functional profiling. Taxonomic assignment of representative OTUs was performed using the RDP Classifier (26) against Greengenes v13.8 database (27). Alignment of sequences was performed using PyNast (28) as default in QIIME pipeline. Chimera checking was performed using Chimera Slayer (29).
We applied two extra filtering steps in aligned and taxonomy-assigned OTU table. First, sequences that belonged to chloroplasts class were filtered out. After that, sequences representing less than 0.005% of total OTUs were also filtered out [as previously done in Ref. (30)] from the chloroplast filtered OTU table. After these two extra filtering steps, we lost a mean of 27% of sequences (median of 25%, ranging from 2 to 77%) and a mean of 21% of sequences (median of 15%, ranging from 1 to 77%) in open and closed refer-ence approaches, respectively (Data Sheet S1 in Supplementary Material).
Downstream Bioinformatics analyses: Diversity, composition, Potential Functions, and statistical TestsDownstream analyses were performed using QIIME 1.9.1 (24) with the filtered OTU table. Reads are clustered against a
reference sequence collection, and all of the reads that do not hit a sequence in the reference sequence collection are excluded from downstream analyses in a closed reference approach or are subsequently clustered de novo in an open reference approach. To standardize samples with unequal sequencing depths, analyses were performed using random subsets of 25,000 sequences per sample in the open reference approach and random subsets of 10,000 sequences per sample in the closed reference approach. The perianal sample of one FB (FB1.8) failed this parameter and was discarded for posterior analyses.
Alpha diversity analysis assesses the diversity within a sample. In alpha diversity, we used two different metrics: observed spe-cies to assess richness and Shannon index to assess evenness. We assessed statistical significance with 999 permutations using the non-parametric Monte Carlo permutation test and corrected the P value through false discovery rate.
To assess the differences in the alpha diversity and composition at the individual level, we collapsed the eight skin samples from a dog using QIIME v1.9.1 to form a unique sample representing the individual. Therefore, the sample size for analyzing the individual effect is nine.
To assess the differences in the alpha diversity and compo-sition when considering the breed, we used two approaches: (A) analyzing each skin site independently and (B) using the QIIME collapsed values from the eight skin site samples for each dog (n = 9). In the first approach, we group the three samples corresponding to a skin site from a breed and assessed differences in breeds per skin site; e.g., GS1.1, GS2.1, and GS3.1 as GS_chin and we compared them to FB_chin and WH_chin. In the second approach, we group the three collapsed individual samples from each breed; e.g., GS1, GS2, and GS3 as GS and we compared them to FB and WH.
Beta diversity analysis assesses the similarities among samples of the same community. Beta diversity was performed using both weighted and unweighted UniFrac distance metrics (31). Weighted UniFrac considers phylogeny, taxa, and relative abundances, whereas unweighted UniFrac only considers phy-logeny and taxa. Those distance matrices were used to create PCoA plots and unweighted pair group method with arithmetic mean (UPGMA) trees. Trees were plotted using FigTree (REF). ANOSIM and adonis statistical methods were applied to evaluate if some variables were determining grouping and to which extent.
PICRUSt (18) was used to predict the functional profile of skin bacterial communities using 16S rRNA gene data obtained using a closed reference approach in QIIME v1.9.1. Kyoto Encyclopedia of Genes and Genomes (KEGG) (32) Ortholog (KO) hierarchy was used to make inferences of the functional gene content.
Linear discriminant analysis (LDA) effect size (LEfSe) (33) was used to compare groups and to identify differentially abundance distribution in both taxa and predicted functions (α = 0.05 and with an LDA score >3.0).
resUlTs
To assess variability and composition of dog skin microbiota, we performed a cross-sectional study with healthy dogs from three breeds. We have analyzed 72 samples from 9 dogs: 3 FBs,
4
Cuscó et al. Skin Microbiota in Healthy Dogs
Frontiers in Veterinary Science | www.frontiersin.org February 2017 | Volume 4 | Article 6
3 GSs, and 3 West Highland WH. We sampled eight skin sites: chin, inner pinna, nasal skin, axilla, back, abdomen, interdigital region, and perianal area, which are named as 1, 2, 3, 4, 5, 6, 7, and 8, respectively (Figure 1). These anatomic sites were selected to represent the regional diversity of the canine skin (34).
We found a total of 2,092 bacterial OTUs living on dog skin, which were taxonomically classified into 20 phyla, 51 classes, 69 orders, 132 families, and 245 genera. Data Sheet S2 in Supplementary Material contains several OTU tables: the complete OTU table for the 72 samples, the OTU table at family level obtained for all the samples, the OTU table collapsed by site, and the OTU table collapsed by individual.
The abundances of the main phyla differed on each sample (Figure 2A). The main phyla on skin samples were Proteo-bacteria (1–73%), Firmicutes (3–93%), Fusobacteria (0–58%), Bacte roidetes (0–69%), and Actinobacteria (0–35%), followed by Cyano bacteria, Tenericutes, TM7, and others with lower abundances.
Alpha diversity values were also very variable among samples (Data Sheet S3 and Figure S1 in Supplementary Material). The richness (observed species) ranged from 145.6 in the chin of WH1 to 928.8 in the inner pinna of GS1 (average of 488.42). The evenness (Shannon Index) ranged from 0.959 in the axilla of WH3 to 8.559 in the abdomen of WH2 (average of 5.8).
To assess if the variability of the dog skin microbiota depended on individual, breed, and/or skin site and to which extent, we clustered the samples using UPGMA trees and assessed statisti-cal significance using adonis and ANOSIM tests (Figures 2B,C). We found that the main force driving the variability in dog skin microbiota composition is the individual, followed by the skin site and the breed.
Despite the high variability detected among samples, all of them were skin microbiota of healthy dogs and in fact shared some of their taxonomy. Thus, to assess the homogeneity of the samples, we analyzed the core microbiota. To complete the analysis, we assessed the potential functions of the bacterial com-munity using PICRUSt.
individualSamples from the same individual tended to cluster together (Figures 2B,C). Statistical analysis using adonis test con-firmed this result: the clustering of samples per individual significantly explained 40% (unweighted UniFrac) and 45% (weighted UniFrac; Figure S2A in Supplementary Material) of the distances among samples. Moreover, ANOSIM R value was close to +1 (R = 0.79, P = 0.001) in unweighted UniFrac, suggesting a strong dissimilarity among groups that was mainly explained by low-abundant species in each dog. Therefore, the individual was the variable that explained most differences among samples.
Seven of nine dogs had a taxonomic profile with the main bacterial phyla: Firmicutes and Proteobacteria with higher abun-dances than Actinobacteria or Bacteroidetes (Figure 2A). From these dogs, FB2, FB3, and WH1 presented also Fusobacteria as one of the main phyla if not the greatest one, whereas in GS1, GS2, GS3, and FB1, this phylum was almost absent. Two of nine dogs presented a predominant phylum (>50% of the total abundance)
over the others, WH2 with Proteobacteria and WH3 with Firmicutes. The abundances of these two phyla and others were differentially distributed (Figure S2B in Supplementary Material) (α = 0.05, LDA score >3).
The abundances differed in each individual, not only at the phylum level but also at the deeper taxonomic levels, such as family level (Data Sheet S2 in Supplementary Material). We can detect some individual-specific families, when looking at the most abundant families (Table 1): Listeriaceae representing a 22.5% of total microbiota composition for GS2; Porphyromonadaceae with a 26.1% for WH1; and or Enterobacteriaceae with a 12% for FB1. On the other hand, Streptococcaceae was present in all the individuals with low percentages, in exception of WH3 with 59% of the total composition that making it the individual with the lowest evenness value (3.71 of Shannon Index, Data Sheet S3 in Supplementary Material). Depending on the individual, families representing more than 5% (Table 1) were describing from 36.3 to 78.6% of total microbiota composition.
skin siteClustering samples per skin site significantly explained 19% of the distances, when considering composition, phylogeny, and relative abundances (weighted UniFrac; Figure 3A). Visually inspecting the beta diversity plot, we found that perianal samples cluster together.
Inner pinna presented the highest diversity value with an average richness of 610.82 observed species and an average evenness of 6.85 of Shannon index, whereas the perianal region presented the lowest diversity, with only 323.1 observed species and a Shannon index of 4.41 (Data Sheet S3 in Supplementary Material). These two skin regions were significantly different between each other when considering evenness (Figure 3B; P = 0.028).
Most of the skin sites had Proteobacteria and Firmicutes as the most abundant phyla, adding up to more than 55% of the total microbiota composition (Figure 3C). Chin had also Bacteroidetes as an abundant phylum with a 14.6% of Porphyromonadaceae, becoming the main family of this skin site. The perianal region was the exception and presented the most different composition profile (Figure 3 and Table 2; Data Sheet S2 in Supplementary Material). In perianal region, Bacteroidetes was the main phylum, followed by Firmicutes and Fusobacteria. Moreover, Proteobacteria, which is one of the main phyla inhabiting dog skin, was almost absent. The three main families inhabiting perianal region were Bacteroidaceae with 32.5% (Bacteroidetes), Fusobacteriaceae with 25.6% (Fusobacteria), and Lachnospiraceae with 6.4% (Firmicutes).
We detected differentially distributed abundances on skin sites with LEfSe analyses (α = 0.05, LDA score >3) at the phylum and class level (Figure 3D) and up to the family level (Figure S3 in Supplementary Material). At the phylum and class level, Proteobacteria was significantly overrepresented at inner pinna, mainly due to the members of Alphaproteobacteria class; and Actinobacteria were overrepresented at the back. Some of the lowest abundant phyla were significantly more represented in a specific skin site: GN02 and TM7 at chin and [Thermi] (mainly from the Deinococci class) at inner pinna.
FigUre 2 | individual signatures on dog skin. (a) Taxonomy bar plot of each sample at phylum level. The first two letters with the number represent the breed and individual; and the numbers represent each skin site. (B) PCoA plot using unweighted UniFrac metrics colored by individual with values of ANOSIM and adonis statistical tests. (c) Unweighted pair group method with arithmetic mean tree associated with (B); branches are colored by individual.
5
Cuscó et al. Skin Microbiota in Healthy Dogs
Frontiers in Veterinary Science | www.frontiersin.org February 2017 | Volume 4 | Article 6
BreedThe breed explained fewer differences among the samples, but it did explain some differences. Clustering samples per breed significantly explained 10% (unweighted UniFrac) and 9%
(weighted UniFrac) of the distances among samples (Figures S4A,B in Supplementary Material).
To assess the effect of the breed in diversity and composition, we used two approaches: (A) analyzing each skin site separately
FigUre 3 | Dog skin microbiota analysis considering site. (a) PCoA plot using weighted UniFrac metrics colored by skin site with values of ANOSIM and adonis statistical tests. Perianal region is circled in brown. (B) Boxplots of alpha diversity values. Marked with a red asterisk the two comparisons that were statistically different when using Monte Carlo permutation test (P < 0.05). (c) Bar plot representing skin microbiome composition at phylum level per skin site; each bar represents the mean values of the nine dogs per each skin site. (D) Histogram of linear discriminant analysis (LDA) effect size scores for differentially abundance distribution (α = 0.05, LDA score >3) of bacterial phyla and classes among individuals.
TaBle 1 | skin microbiota composition at family level for each individual.
Phylum Family FB1 FB2 FB3 gs1 gs2 gs3 Wh1 Wh2 Wh3
Proteobacteria Rhodospirillaceae 0.4% 1.5% 1.7% 2.4% 1.2% 5.5% 0.2% 10.9% 0.5%Proteobacteria Sphingomonadaceae 2.5% 3.3% 1.5% 3.0% 1.2% 1.9% 0.7% 9.0% 0.7%Proteobacteria Enterobacteriaceae 12.0% 0.2% 0.1% 0.2% 0.3% 0.1% 0.1% 0.0% 0.2%Proteobacteria Pasteurellaceae 6.4% 1.2% 0.1% 0.2% 0.9% 0.4% 0.4% 0.1% 0.2%Proteobacteria Moraxellaceae 0.8% 0.2% 0.3% 0.3% 2.0% 0.4% 0.5% 6.6% 0.4%Firmicutes Listeriaceae 0.1% 0.0% 0.0% 0.1% 22.5% 0.5% 0.2% 0.1% 0.0%Firmicutes Staphylococcaceae 10.2% 4.6% 7.0% 5.6% 1.2% 10.6% 0.8% 3.9% 0.3%Firmicutes Streptococcaceae 2.2% 2.6% 1.1% 1.0% 3.3% 1.6% 0.2% 0.1% 59.1%Firmicutes Clostridiaceae 0.5% 0.5% 8.2% 4.9% 1.4% 1.6% 0.9% 0.4% 2.6%Firmicutes Lachnospiraceae 0.4% 0.3% 6.9% 1.8% 0.6% 1.0% 2.9% 0.9% 0.7%Fusobacteria Fusobacteriaceae 0.6% 21.2% 23.9% 1.9% 2.3% 5.4% 32.2% 3.4% 3.6%Fusobacteria Leptotrichiaceae 0.3% 5.9% 0.0% 0.1% 0.1% 0.1% 0.0% 0.0% 0.0%Bacteroidetes Bacteroidaceae 0.4% 3.1% 6.6% 6.2% 9.0% 8.6% 1.4% 5.7% 0.3%Bacteroidetes Porphyromonadaceae 1.8% 1.6% 3.3% 0.2% 4.7% 1.0% 26.1% 1.4% 8.5%Bacteroidetes Weeksellaceae 5.4% 0.8% 0.3% 0.8% 3.5% 1.8% 1.7% 1.0% 0.2%Actinobacteria Corynebacteriaceae 4.8% 0.5% 1.4% 6.7% 1.0% 2.6% 2.1% 1.4% 1.3%Actinobacteria Intrasporangiaceae 5.2% 1.4% 0.4% 0.9% 0.3% 2.6% 2.5% 0.3% 0.0%
% of microbiota explained by taxa >5% 53.9% 48.9% 62.8% 36.3% 55.6% 45.6% 72.6% 45.0% 78.6%
List of the taxa that represent >5% of the total microbiota composition and their abundances considering the individual.
6
Cuscó et al. Skin Microbiota in Healthy Dogs
Frontiers in Veterinary Science | www.frontiersin.org February 2017 | Volume 4 | Article 6
TaBle 3 | skin core microbiota at family level for each individual and skin site.
Phylum Family gs1 gs2 gs3 FB1 FB2 FB3 Wh1 Wh2 Wh3
Actinobacteria Corynebacteriaceae 6.7% 1.0% 2.6% 4.8% 0.5% 1.4% 2.1% 1.4% 1.3%Firmicutes Streptococcaceae 1.0% 3.3% 1.6% 2.2% 2.6% 1.1% 0.2% 0.1% 59.1%
Lachnospiraceae 1.8% 0.6% 1.0% 0.4% 0.3% 6.9% 2.9% 0.9% 0.7%Fusobacteria Fusobacteriaceae 1.9% 2.3% 5.4% 0.6% 21.2% 23.9% 32.2% 3.4% 3.6%Proteobacteria Comamonadaceae 3.3% 1.5% 1.8% 1.6% 1.1% 0.6% 0.9% 1.9% 0.8%
Oxalobacteraceae 2.8% 3.4% 2.2% 1.1% 0.9% 0.5% 0.6% 3.5% 3.9% Neisseriaceae 1.6% 3.9% 1.5% 4.6% 0.8% 1.0% 1.9% 4.9% 0.4%
% of microbiota explained by core taxa 19.1% 16.1% 16.2% 15.3% 27.5% 35.5% 40.7% 16.0% 69.9%
Phylum Family chin inner pinna
nasal skin
axilla Back abdomen interdigital Perianal
Actinobacteria Corynebacteriaceae 6.2% 2.6% 1.2% 2.0% 1.8% 1.6% 0.9% 3.1%Firmicutes Streptococcaceae 5.2% 4.3% 10.6% 11.1% 6.2% 9.5% 11.0% 5.8%
Lachnospiraceae 1.1% 0.7% 0.6% 0.8% 1.0% 0.7% 3.0% 6.4%Fusobacteria Fusobacteriaceae 7.8% 6.7% 11.6% 10.6% 7.4% 13.5% 3.5% 25.6%Proteobacteria Comamonadaceae 1.7% 1.4% 1.9% 2.0% 1.9% 1.8% 1.2% 0.1%
Oxalobacteraceae 1.5% 3.2% 1.6% 2.0% 2.6% 2.1% 3.6% 0.2% Neisseriaceae 4.9% 1.3% 3.3% 1.5% 0.6% 1.1% 5.1% 0.0%
% of microbiota explained by core taxa 28.4% 20.2% 30.8% 30.0% 21.4% 30.4% 28.2% 41.2%
List of the taxa shared in all samples included in the study, their abundances, and distributions by individual and skin site.
TaBle 2 | skin microbiota composition at family level for each skin site.
Phylum Family chin inner pinna nasal skin axilla Back abdomen interdigital Perianal
Proteobacteria Rhodospirillaceae 2.3% 6.6% 2.9% 3.1% 3.0% 0.8% 2.9% 0.1%Proteobacteria Neisseriaceae 4.9% 1.3% 3.3% 1.5% 0.6% 1.1% 5.1% 0.0%Proteobacteria Enterobacteriaceae 0.3% 6.7% 2.1% 0.6% 0.2% 0.1% 0.1% 0.5%Proteobacteria Moraxellaceae 1.0% 0.2% 6.7% 0.6% 0.5% 0.6% 0.3% 0.1%Firmicutes Listeriaceae 1.7% 2.9% 0.2% 6.9% 5.7% 2.7% 0.8% 0.1%Firmicutes Staphylococcaceae 4.5% 2.2% 12.9% 1.7% 5.1% 9.6% 1.6% 0.5%Firmicutes Streptococcaceae 5.2% 4.3% 10.6% 11.1% 6.2% 9.5% 11.0% 5.8%Firmicutes Clostridiaceae 1.6% 2.6% 1.1% 0.8% 1.5% 1.4% 5.4% 4.6%Firmicutes Lachnospiraceae 1.1% 0.7% 0.6% 0.8% 1.0% 0.7% 3.0% 6.4%Fusobacteria Fusobacteriaceae 7.8% 6.7% 11.6% 10.6% 7.4% 13.5% 3.5% 25.6%Bacteroidetes Bacteroidaceae 0.4% 0.2% 0.4% 3.1% 0.6% 2.5% 0.5% 32.5%Bacteroidetes Porphyromonadaceae 14.6% 3.3% 7.3% 5.2% 4.4% 2.8% 4.3% 1.3%Bacteroidetes Weeksellaceae 6.0% 0.8% 2.4% 0.8% 0.6% 1.1% 1.5% 0.0%Actinobacteria Corynebacteriaceae 6.2% 2.6% 1.2% 2.0% 1.8% 1.6% 0.9% 3.1%
% of microbiota explained by taxa >5% 57.5% 41.2% 63.3% 48.8% 38.3% 48.1% 40.8% 80.6%
List of the taxa that represent >5% of the total microbiota composition and their abundances considering the skin site.
7
Cuscó et al. Skin Microbiota in Healthy Dogs
Frontiers in Veterinary Science | www.frontiersin.org February 2017 | Volume 4 | Article 6
considering the breed and (B) analyzing each collapsed dog sample per breed, adding up together all the values of the eight skin sites to form an individual dog value and grouping the three dogs from the same breed.
At taxonomic composition level, when analyzing each skin site per breed, we saw some differences (α = 0.05, LDA score >3) (Figure S4C in Supplementary Material). At phylum level, Tenericutes were overrepresented at nasal skin of FB. At family level, GS had an overrepresentation of Dermabacteraceae at axilla and Corynebacteriaceae and Williamsiaceae at the interdigital region, whereas FB had Burkholderiaceae and Bacillaceae at axilla, Gemellaceae at the interdigital region, and Gordoniaceae at back and chin. When collapsing all the eight skin sites to obtain an individual sample, we only detected three families with differentially distributed abundances: Sphingobacteriaceae and Dermabacteraceae in GS and Enterococcaceae in FB (Figure
S4D in Supplementary Material). All of these taxa had low relative abundances (Data Sheet S2 in Supplementary Material).
In alpha diversity analysis, we detected no statistical differ-ences, both when analyzing each skin site separately (Figure S5A in Supplementary Material) and when analyzing the collapsed dog samples (Figure S5B in Supplementary Material).
core skin MicrobiotaEach dog had its own microbiota profile, but there were also taxa shared among all samples even at low-abundant level, which we can define as the skin core microbiota of our cohort of individuals.
Families found in all the skin samples analyzed in this study were Corynebacteriaceae (Actinobacteria); Streptococcaceae and Lachnospiraceae (Firmicutes); Fusobacteriaceae (Fusobacteria); and Comamonadaceae, Oxalobacteraceae, and Neisseriaceae (Proteobacteria) (Table 3).
8
Cuscó et al. Skin Microbiota in Healthy Dogs
Frontiers in Veterinary Science | www.frontiersin.org February 2017 | Volume 4 | Article 6
The skin core microbiota at family level explained from 15.3 to 40.7% of the individual composition and from 20.2 to 41.2% of the skin site composition. It reached 69.9% for WH3 (with 59.1% of Streptococcaceae). Although a group of families constituted the core microbiota, their abundances were specific for each individual and site.
When we consider that the skin core microbiota is defined by taxa present in 85% of the samples (61 of 71 samples; to exclude some specific site or specific individual), the skin core microbiota explained a mean of 78% of the skin composition at both the indi-vidual and the skin site level, and we found 39 different families (Data Sheet S4 in Supplementary Material).
Predicted FunctionsWe used 16S rRNA gene sequencing data to predict the func-tional profile of dog skin microbiota samples, applying PICRUSt. PICRUSt developers (18) and more recently Meisel et al. (35) reported strong correlations between human metagenomic data sets and 16S-based functional prediction in skin microbiota.
We found up to 41 predicted functions for the dog skin microbiota, when considering the second level of KO hierarchy. Membrane transport (environmental information processing); replication and repair (genetic information processing); and amino acid, carbohydrate, and energy metabolism (metabolism) are the functions more spread and represented, with a mean rela-tive abundance of 12, 7.9, 10.3, 10.4, and 5.6%, respectively (Data Sheet S6 in Supplementary Material).
Taxa composition profiles became more uniform when con-verting them to predicted functions (Figures 4A,B). However, we found some differentially distributed abundances in predicted functions at breed, individual, and skin site level (α = 0.05, LDA score >3). We focused on assessing differences on the functional prediction among skin sites (Figure 4C).
Some predicted functions were overrepresented in back, chin, perianal region, and inner pinna and differentially distributed from all other sites. In pinna, we found overrepresentation of cellular processes and cell motility (cellular processes) and also signal transduction (environmental information processing). In perianal region, three metabolism pathways were increased: car-bohydrate metabolism, glycan biosynthesis and metabolism, and nucleotide metabolism. In chin, we found overrepresentation of genetic information processing and its sublevel pathways—repli-cation and repair and translation. In the back, three metabolism pathways were increased: xenobiotics biodegradation and metabolism, lipid metabolism, and metabolism of terpenoids and polyketides. Figure S6 in Supplementary Material contains LEfSe plots of differentially abundant predicted functions at level 3 of KEGG Orthology for skin site.
DiscUssiOn
Our results suggest that the main force driving the variability in microbiota composition in dogs is the individual, rather than the breed—hair coat—or the skin site. This is true both considering the community structure (weighted UniFrac), but mainly when looking at the less abundant species (unweighted UniFrac). Several human studies have reported that interindividual
variation is high and defines a “personal microbiome” (9, 19, 36). These low abundant bacterial signatures have been even used to identify individuals (37).
Meason et al. found recently this same pattern for canine skin mycobiome (fungal community) (17). Moreover, Rodrigues-Hoffmann et al. observed great differences on individuals, although they focused on detecting skin site differences and not on assessing the effect of the individual directly (6). On the other hand, human skin has three main microhabitats or ecological niches, depending on the physiological properties: sebaceous, dry, and moist areas and different microbiota is associated with each microhabitat (5). Conversely, dogs present almost all their skin sites covered by hair that creating a more uniform habitat.
Previous research had detected Proteobacteria (6, 7) or Firmicutes (38) as the main phyla inhabiting dog skin microbiota. Our results suggest that either Proteobacteria or Firmicutes or a combination of both can be the main phyla, depending on the individual. We also found Fusobacteria as one of the most abundant phyla for three of nine dogs, and when it was present, it spread over all the skin sites. Rodrigues-Hoffmann et al. detected Fusobacteria as one phylum specific to perianal regions (6); other studies also found them in groins (8) and paws and forehead (7), but with lower abundances than those seen here.
At the family level, taxa found in our cohort resemble those found in other canine skin microbiota studies (6–8, 18). Rodrigues-Hoffmann et al. found that Oxalobacteraceae, spe-cifically Ralstonia spp., was the most abundant and extended taxa on dog skin (6), specially on healthy dogs; however, none of our Oxalobacteraceae sequences were from Ralstonia spp. Pierezan et al. have suggested that this could be due to the use of different supplies for the collection of samples, modifications in sample storage, extraction methods, and/or changes in the high-throughput sequencing platform used (8). Ralstonia spp. had been also detected in “blank” controls in microbiota studies and could be contaminants from the laboratory or the kits and reagents used (39).
Among all the individuals included, WH3 was very different with its skin mostly inhabited by Streptococcaceae that suggest-ing a colonization event. The representative sequence of the most abundant Streptococcaceae OTU in WH3 corresponds to Streptococcus canis, which are considered opportunistic patho-gens inhabiting healthy dog skin. Their overgrowth has been associated to dermatitis (40) and even necrotizing fasciitis (41). Moreover, WH3 was the less diverse individual. Low alpha diver-sity values were characterizing skin microbiota in dogs affected by atopic dermatitis (6, 18), and in humans, they had been linked to elderly people (42). Therefore, we have two hypotheses for WH3: although considered healthy by the clinicians, the dog was beginning to develop some skin affliction; or the effect could be due to its advanced age. Further studies would be needed to assess the effect of age on healthy dog skin microbiota. Reanalyzing results excluding this sample have shown similar results for both ANOSIM and adonis tests (data not shown), confirming that the individual is the main force driving microbiota structure and composition and that the inclusion of this dog does not interfere with the results obtained.
FigUre 4 | skin microbiota relative abundances: taxa vs predicted functions. Bar plots obtained through a closed reference approach to be comparable between them (see Section “Materials and Methods”). Each stacked bar represents relative abundances of each sample included in the study. Relative abundances of (a) bacteria at phylum level and (B) predicted functions (second level of the Kyoto Encyclopedia of Genes and Genomes Ortholog hierarchy) based on Phylogenetic Investigation of Communities by Reconstruction of Unobserved States data set. (c) Histogram of linear discriminant analysis (LDA) effect size scores for differentially abundance distribution (α = 0.05, LDA score >3) of predicted functions. Complete list of predicted functions on dog skin and their relative abundances is available in Data Sheet S6 in Supplementary Material.
9
Cuscó et al. Skin Microbiota in Healthy Dogs
Frontiers in Veterinary Science | www.frontiersin.org February 2017 | Volume 4 | Article 6
10
Cuscó et al. Skin Microbiota in Healthy Dogs
Frontiers in Veterinary Science | www.frontiersin.org February 2017 | Volume 4 | Article 6
Despite the major force driving microbiota composition and variability was the individual, skin site also plays a role explaining the variability observed. The variability of the skin microbiota regarding the site could be due to the influences of other body site microbiota, such as the perianal region with the gastrointestinal microbiota or the chin with the oral microbiota, or due to the specific physiological properties of each skin site, such as the back with higher sebum production. Perianal region presented the most different composition profile: Bacteroidetes followed by Firmicutes and Fusobacteria were the main phyla, whereas Proteobacteria presented lower abundances. Moreover, Erysipelotrichi and Clostridia classes were overrepresented. This phyla pattern and taxa are more similar to that seen on canine gastrointestinal microbiota than that from the skin (20, 43). At the functional level profiling, some metabolic pathways were significantly overrepresented in the perianal region. Swanson et al. detected carbohydrate metabolism as one of the main pathways in intestinal microbiota of dogs, with values similar to those detected here (20) that are differentially higher than the other skin sites included. In chin, the most abundant phyla were Bacteroidetes, followed by Proteobacteria and Firmicutes. Sturgeon et al. have detected those same three phyla as the most abundant ones on canine oral microbiota (44). Moreover GN02 and TM7, two of the lowest abundant phyla, were overrepresented and differentially distributed on that region. These two phyla have been previously detected in canine oral microbiota (45). We also found that Porphyromonadaceae and Fusobacteraceae are the most abundant families in chin, coinciding with Bradley et al. who detected Porphyromonas and Fusobacterium (among others) as abundant genera in canine oral microbiome (16). On the other hand, physiological properties of the back skin could be influencing the microbiota function of that region. The dorsal parts of the neck, the trunk, and the tail have larger sebaceous glands than other skin regions (46). Moreover, the dorsal region has the densest hair coat, so a larger number of sebaceous glands associated with the hair follicles (46). Consequently, more sebum is produced than in other skin sites, which is mainly composed of lipid compounds. The higher abundance of this substrate could be explaining the increased lipid metabolism and fatty acid metabolism pathways in microbiota inhabiting back.
Even when our results show that the main force driving skin microbiota structure and composition is the individual, we cannot rule out the influence of the environment and lifestyle. The individual should be understood as the dog, its lifestyle, and its environment. In fact, the chloroplasts sequences that we detected and discarded for the ulterior analysis were not evenly distributed, but more represented in three dogs (Data Sheet S1 in Supplementary Material), suggesting that these individuals had a greater or more recent exposure to outdoor environment and may have more transient bacterial members detected as skin microbiota. On the other hand, despite being the human skin constantly exposed to extrinsic factors, healthy adults have shown to maintain their skin microbial communities over time (36). This last hypothesis should be assessed in dogs, because they are exposed to extrinsic factors, such as environment or
human contact. In this study, we cannot distinguish whether this individual factor is solely host specific or it also includes extrinsic properties from the environment.
Some of the differences when comparing our results to previous studies could be due to differences in the methodolo-gies chosen such as the 16S region analyzed or the sequencing platform used. We are amplifying 400 bp of the V1–V2 hyper-variable regions that had been suggested to be a better choice for skin microbiota in humans among others (47). Hypervariable regions V1–V3 are the most commonly used on dog skin microbiota studies (6, 18), but only V2 region has also been used (7). Recently, Pierezan et al. used V4 (8). On the other hand, Clooney et al. found that the factor responsible for the greatest variance in microbiota composition was the chosen methodology, when comparing Illumina HiSeq, Illumina MiSeq, and Ion Torrent PGM. This problem was larger in Illumina MiSeq rather than in Ion Torrent PGM when analyz-ing 16S rRNA V1–V2 region amplicons (48). In another study comparing microbial profiles using V1–V2 regions, the authors concluded that the output generated from PGM Ion Torrent and 454 yielded concurrent results (49). Finally, PICRUSt is a tool that was mainly developed for the human microbiome. However, dogs share skin microbiota with their owners [as seen in Ref. (7)]. So, using PICRUSt for skin in pets is probably a valid approach. Moreover, PICRUSt has already been used in fecal samples of dogs (21).
cOnclUsiOn
The individual seems to be the main force driving skin micro-biota composition and diversity in dogs, and dissimilarity is mainly explained by low-abundant species in each dog. The main phyla inhabiting the dog skin in our cohort are Proteobacteria, Firmicutes, Fusobacteria, Actinobacteria, and Bacteroidetes, and their abundance patterns differ among individuals.
The skin site also plays a role: the composition and function of microorganisms inhabiting chin and perianal region could be influenced by other body site microbiota. Moreover, the specific physiological properties of the back, with higher abundance of sebum, could favor the growth of specific microorganisms. We observed distinctive taxa composition profiles for each sample, but relative abundances become more uniform when converting them to predicted functions.
As the diversity among individuals is the highest, a good choice to better assess the dog skin microbiota would probably be comparing affected vs unaffected regions from the same dog rather than comparing different dogs in case–control studies, so each dog is its own control; and an accurate assessment of the environmental factor, controlling variables such as geographical region, season, lifestyle, or cohabitation with other animals.
aUThOr cOnTriBUTiOns
AS, LF, and OF conceived and designed the experiment. AS, LA, LF, and OF supervised the project and gave conceptual advice. AC and OF performed the experiment. AC carried out the
11
Cuscó et al. Skin Microbiota in Healthy Dogs
Frontiers in Veterinary Science | www.frontiersin.org February 2017 | Volume 4 | Article 6
bioinformatics analysis. AC drafted the manuscript. AS, LA, LF, and OF edited the manuscript. All authors read and approved the final manuscript.
acKnOWleDgMenTs
We acknowledge Marc Pons (UAB) for helping at the initial bioinformatics analysis and Diana Ferreira, Xavier Roura, and Mar Bardagí (Hospital Clinic Veterinari, UAB) for the sample
collection. This work was supported by a grant awarded by Generalitat de Catalunya (Industrial Doctorate program, 2013 DI 011).
sUPPleMenTarY MaTerial
The Supplementary Material for this article can be found online at http://journal.frontiersin.org/article/10.3389/fvets.2017.00006/full#supplementary-material.
reFerences
1. Grice EA, Kong HH, Renaud G, Young AC, Bouffard GG, Blakesley RW, et al. A diversity profile of the human skin microbiota. Genome Res (2008) 18:1043–50. doi:10.1101/gr.075549.107
2. Naik S, Bouladoux N, Wilhelm C, Molloy MJ, Salcedo R, Kastenmuller W, et al. Compartmentalized control of skin immunity by resident commensals. Science (2012) 337:1115–9. doi:10.1126/science.1225152
3. Oh J, Byrd AL, Deming C, Conlan S, Barnabas B, Blakesley R, et al. Biogeography and individuality shape function in the human skin metage-nome. Nature (2014) 514:59–64. doi:10.1038/nature13786
4. Grice EA, Segre JA. The skin microbiome. Nat Rev Microbiol (2011) 9:244–53. doi:10.1038/nrmicro2537
5. Grice EA, Kong HH, Conlan S, Deming CB, Davis J, Young AC, et al. Topographical and temporal diversity of the human skin microbiome. Science (2009) 324:1190–2. doi:10.1126/science.1171700
6. Rodrigues-Hoffmann A, Patterson AP, Diesel A, Lawhon SD, Ly HJ, Elkins Stephenson C, et al. The skin microbiome in healthy and allergic dogs. PLoS One (2014) 9:e83197. doi:10.1371/journal.pone.0083197
7. Song SJ, Lauber C, Costello EK, Lozupone CA, Humphrey G, Berg-Lyons D, et al. Cohabiting family members share microbiota with one another and with their dogs. Elife (2013) 2:1–22. doi:10.7554/eLife.00458
8. Pierezan F, Olivry T, Paps JS, Lawhon SD, Wu J, Steiner JM, et al. The skin microbiome in allergen-induced canine atopic dermatitis. Vet Dermatol (2016) 5:332–e82. doi:10.1111/vde.12366
9. Costello EK, Lauber CL, Hamady M, Fierer N, Gordon JI, Knight R. Bacterial community variation in human body habitats across space and time. Science (2009) 326:1694–7. doi:10.1126/science.1177486
10. Findley K, Grice EA. The skin microbiome: a focus on pathogens and their association with skin disease. PLoS Pathog (2014) 10:e1004436. doi:10.1371/journal.ppat.1004436
11. Kong HH, Oh J, Deming C, Conlan S, Grice EA, Beatson MA, et al. Temporal shifts in the skin microbiome associated with disease flares and treatment in children with atopic dermatitis. Genome Res (2012) 22:850–9. doi:10.1101/gr.131029.111
12. Oh J, Freeman AF, Park M, Sokolic R, Candotti F, Holland SM, et al. The altered landscape of the human skin microbiome in patients with primary immunodeficiencies. Genome Res (2013) 23:2103–14. doi:10.1101/gr. 159467.113
13. Alekseyenko AV, Perez-Perez GI, De Souza A, Strober B, Gao Z, Bihan M, et al. Community differentiation of the cutaneous microbiota in psoriasis. Microbiome (2013) 1:31. doi:10.1186/2049-2618-1-31
14. Takemoto A, Cho O, Morohoshi Y, Sugita T, Muto M. Molecular character-ization of the skin fungal microbiome in patients with psoriasis. J Dermatol (2015) 42:166–70. doi:10.1111/1346-8138.12739
15. Fitz-Gibbon S, Tomida S, Chiu BH, Nguyen L, Du C, Liu M, et al. Propionibacterium acnes strain populations in the human skin microbiome associated with acne. J Invest Dermatol (2013) 133:2152–60. doi:10.1038/jid.2013.21
16. Bradley CW, Morris DO, Rankin SC, Cain CL, Misic AM, Houser T, et al. Longitudinal evaluation of the skin microbiome and association with micro-environment and treatment in canine atopic dermatitis. J Invest Dermatol (2016) 136:1182–90. doi:10.1016/j.jid.2016.01.023
17. Meason-smith C, Diesel A, Patterson AP, Older CE, Mansell JM, Suchodolski JS, et al. What is living on your dog’s skin? Characterization of the canine cutaneous mycobiota and fungal dysbiosis in canine
allergic dermatitis. FEMS Microbiol Ecol (2015) 91:1–12. doi:10.1093/femsec/ fiv139
18. Langille MGI, Zaneveld J, Caporaso JG, McDonald D, Knights D, Reyes JA, et al. Predictive functional profiling of microbial communities using 16S rRNA marker gene sequences. Nat Biotechnol (2013) 31:814–21. doi:10.1038/nbt.2676
19. Human Microbiome Project Consortium. Structure, function and diversity of the healthy human microbiome. Nature (2012) 486:207–14. doi:10.1038/nature11234
20. Swanson KS, Dowd SE, Suchodolski JS, Middelbos IS, Vester BM, Barry KA, et al. Phylogenetic and gene-centric metagenomics of the canine intes-tinal microbiome reveals similarities with humans and mice. ISME J (2011) 5:639–49. doi:10.1038/ismej.2010.162
21. Minamoto Y, Otoni CC, Steelman SM, Büyükleblebici O, Steiner JM, Jergens AE, et al. Alteration of the fecal microbiota and serum metabolite profiles in dogs with idiopathic inflammatory bowel disease. Gut Microbes (2015) 6:33–47. doi:10.1080/19490976.2014.997612
22. Cadieu E, Neff MW, Quignon P, Walsh K, Chase K, Parker HG, et al. Coat variation in the domestic dog is governed by variants in three genes. Science (2009) 326:150–3. doi:10.1126/science.1177808
23. Jaeger K, Linek M, Power HT, Bettenay SV, Zabel S, Rosychuk RA, et al. Mueller RS. Breed and site predispositions of dogs with atopic dermatitis: a comparison of five locations in three continents. Vet Dermatol (2010) 21:118–22. doi:10.1111/j.1365-3164.2009.00845.x
24. Caporaso JG, Kuczynski J, Stombaugh J, Bittinger K, Bushman FD, Costello EK, et al. QIIME allows analysis of high-throughput community sequencing data. Nat Methods (2010) 7:335–6. doi:10.1038/nmeth.f.303
25. Edgar RC. Search and clustering orders of magnitude faster than BLAST. Bioinformatics (2010) 26:2460–1. doi:10.1093/bioinformatics/btq461
26. Wang Q, Garrity GM, Tiedje JM, Cole JR. Naive Bayesian classifier for rapid assignment of rRNA sequences into the new bacterial taxonomy. Appl Environ Microbiol (2007) 73:5261–7. doi:10.1128/AEM.00062-07
27. DeSantis TZ, Hugenholtz P, Larsen N, Rojas M, Brodie EL, Keller K, et al. Greengenes, a chimera-checked 16S rRNA gene database and workbench compatible with ARB. Appl Environ Microbiol (2006) 72:5069–72. doi:10.1128/AEM.03006-05
28. Caporaso JG, Bittinger K, Bushman FD, Desantis TZ, Andersen GL, Knight R. PyNAST: a flexible tool for aligning sequences to a template alignment. Bioinformatics (2010) 26:266–7. doi:10.1093/bioinformatics/ btp636
29. Haas BJ, Gevers D, Earl AM, Feldgarden M, Ward DV, Giannoukos G, et al. Chimeric 16S rRNA sequence formation and detection in Sanger and 454-pyrosequenced PCR amplicons. Genome Res (2011) 21:494–504. doi:10.1101/gr.112730.110
30. Navas-molina JA, Peralta-sánchez JM, González A, McMurdie PJ, Vázquez-baeza Y, Xu Z, et al. Advancing our understanding of the human microbi-ome using QIIME. Methods Enzymol (2013) 531:371–444. doi:10.1016/B978-0-12-407863-5.00019-8
31. Lozupone C, Knight R. UniFrac: a new phylogenetic method for compar-ing microbial communities. Appl Environ Microbiol (2005) 71:8228–35. doi:10.1128/AEM.71.12.8228
32. Kanehisa M, Goto S. KEGG: Kyoto encyclopaedia of genes and genomes. Nucleic Acids Res (2000) 28:27–30. doi:10.1093/nar/28.1.27
33. Segata N, Izard J, Waldron L, Gevers D, Miropolsky L, Garrett WS, et al. Metagenomic biomarker discovery and explanation. Genome Biol (2011) 12:R60. doi:10.1186/gb-2011-12-6-r60
12
Cuscó et al. Skin Microbiota in Healthy Dogs
Frontiers in Veterinary Science | www.frontiersin.org February 2017 | Volume 4 | Article 6
34. Miller WH, Griffin CE, Campbell KL, Muller GH. Muller and Kirk’s Small Animal Dermatology. 7th ed. St Louis: Elsevier (2013).
35. Meisel JS, Hannigan GD, Tyldsley AS, Sanmiguel AJ, Hodkinson BP, Zheng Q, et al. Skin microbiome surveys are strongly influenced by exper-imental design. J Invest Dermatol (2016) 136:947–56. doi:10.1016/j.jid. 2016.01.016
36. Oh J, Byrd AL, Park M; NISC Comparative Sequencing Program, Kong HH, Segre JA. Temporal stability of the human skin microbiome. Cell (2016) 165:854–66. doi:10.1016/j.cell.2016.04.008
37. Fierer N, Lauber CL, Zhou N, McDonald D, Costello EK, Knight R. Forensic identification using skin bacterial communities. Proc Natl Acad Sci U S A (2010) 107:6477–81. doi:10.1073/pnas.1000162107
38. Weese JS. The canine and feline skin microbiome in health and disease. Vet Dermatol (2013) 24:137–46. doi:10.1111/j.1365-3164.2012.01076.x
39. Salter SJ, Cox MJ, Turek EM, Calus ST, Cookson WO, Moffatt MF, et al. Reagent and laboratory contamination can critically impact sequence-based microbiome analyses. BMC Biol (2014) 12:87. doi:10.1186/s12915-014-0087-z
40. Lamm CG, Ferguson AC, Lehenbauer TW, Love BC. Streptococcal infection in dogs: a retrospective study of 393 cases. Vet Pathol (2010) 47:387–95. doi:10.1177/0300985809359601
41. DeWinter LM, Low DE, Prescott JF. Virulence of Streptococcus canis from canine streptococcal toxic shock syndrome and necrotizing fasciitis. Vet Microbiol (1999) 70:95–110. doi:10.1016/S0378-1135(99)00128-5
42. Ying S, Zeng D, Chi L, Tan Y, Galzote C, Cardona C, et al. The influence of age and gender on skin-associated microbial communities in urban and rural human populations. PLoS One (2015) 10:e0141842. doi:10.1371/journal.pone.0141842
43. Middelbos IS, Vester Boler BM, Qu A, White BA, Swanson KS, Fahey GC. Phylogenetic characterization of fecal microbial communities of dogs fed diets with or without supplemental dietary fiber using 454 pyrosequencing. PLoS One (2010) 5:e9768. doi:10.1371/journal.pone.0009768
44. Sturgeon A, Stull JW, Costa MC, Weese JS. Metagenomic analysis of the canine oral cavity as revealed by high-throughput pyrosequencing of the 16S rRNA gene. Vet Microbiol (2013) 162:891–8. doi:10.1016/j.vetmic.2012. 11.018
45. Camanocha A, Dewhirst FE. Host-associated bacterial taxa from Chlorobi, Chloroflexi, GN02, Synergistetes, SR1, TM7, and WPS-2 Phyla/candidate divisions. J Oral Microbiol (2014) 6:25468. doi:10.3402/jom.v6.25468
46. Al-Bagdadi F. The integument. In: Evans HE, de Lahunta A, editors. Miller’s Anatomy of the Dog. St Louis: Elsevier (2012). p. 61–79.
47. Kuczynski J, Lauber CL, Walters WA, Parfrey LW, Clemente JC, Gevers D, et al. Experimental and analytical tools for studying the human microbiome. Nat Rev Genet (2011) 13:47–58. doi:10.1038/nrg3129
48. Clooney AG, Fouhy F, Sleator RD, Driscoll AO, Stanton C, Cotter PD, et al. Comparing apples and oranges? Next generation sequencing and its impact on microbiome analysis. PLoS One (2016) 11:e0148028. doi:10.1371/journal.pone.0148028
49. Indugu N, Bittinger K, Kumar S, Vecchiarelli B, Pitta D. A comparison of rumen microbial profiles in dairy cows as retrieved by 454 Roche and Ion Torrent (PGM) sequencing platforms. PeerJ (2016) 4:e1599. doi:10.7717/peerj.1599
Conflict of Interest Statement: The authors declare that the research was con-ducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Copyright © 2017 Cuscó, Sánchez, Altet, Ferrer and Francino. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
81
3.2.1. Annex 1. Erratum: Figure 3
In the production step, an older version of Figure 3 was published instead of the newer.
The published Figure 3 is not incorrect, it only does not match with the legend or text
associated.
Figure 3. Dog skin microbiota analysis considering site. (A) PCoA plot using weighted UniFrac metrics
colored by skin site with values of ANOSIM and adonis statistical tests. Perianal region is circled in brown.
(B) Boxplots of alpha diversity values. Marked with a red asterisk the two comparisons that were statistically
different when using Monte Carlo permutation test (P < 0.05). (C) Bar plot representing skin microbiome
composition at phylum level per skin site; each bar represents the mean values of the nine dogs per each skin
site. (D) Histogram of linear discriminant analysis (LDA) effect size scores for differentially abundance
distribution (α = 0.05, LDA score >3) of bacterial phyla and classes among individuals.
82
3.3. Individual signatures and environmental
factors shape skin microbiota on healthy
dogs
This chapter consists of the article entitled “Individual signatures and environmental
factors shape skin microbiota on healthy dogs” recently submitted to Microbiome journal.
Supplementary material of this article is available online at the following link:
https://www.dropbox.com/sh/kcd1bo4adzlh439/AADsgFQ5MwNYTieimIITZUUAa?dl=0
Individual signatures and environmental
factors shape skin microbiota on healthy dogs
Anna Cuscó1,2*, Janelle M. Belanger3, Liza Gershony3, Alma Islas-Trejo3, Kerinne Levy4,
Juan F. Medrano3, Armand Sánchez1, Anita M. Oberbauer3, Olga Francino1.
1Molecular Genetics Veterinary Service (SVGM), Veterinary School, Universitat Autònoma de Barcelona,
Barcelona, Spain, 2Vetgenomics, Ed Eureka, Parc de Recerca UAB, Barcelona, Spain, 3Department of Animal Science, University of California, Davis, CA, USA, 4Canine Companions for Independence, Santa Rosa, CA, USA.
Abstract
Background. The individual, together with its environment, has been reported as the main
force driving composition and structure of skin microbiota in healthy dogs. Therefore, one
of the major concerns when analyzing canine skin microbiota is the likely influence of the
environment. Despite the dense fur covering, certain skin diseases exhibit differential
prevalence among skin sites, dog breeds and individuals.
Results. Here we aimed to assess the variability of the skin microbiota in healthy dogs
cohabiting together, by analyzing eight different skin sites in a large and homogeneous
cohort of Golden-Labrador Retriever crossbred dogs (N=35). We found that microbiota
composition was driven by the individual, but when considering abundances, the
microbiota structure was driven both by the individual and by the skin site. Network
analyses elucidated bacterial interactions within and between skin sites, especially in chin,
abdomen, axilla and the perianal region, with the highly shared interactions probably
representing an environmental component. When analyzing each skin site independently to
assess host-specific factors, we found that season of birth or the time spent in the kennel
was shaping skin microbiota in all skin sites. The most abundant taxon driving this
difference was Sphingomonas, which is an air-borne bacterium that cannot be cultivated at
elevated temperatures. We also found some taxonomic differences linked to sex on
abdomen, axilla and back. Finally, the USA and European cohorts were grouping by
geographical origin in two different and well-defined clusters, even when the European
individuals were very heterogeneous.
Conclusions. We observed a large inter-individual variability and effects of different host
variables, such as season of birth or time spent in the kennel and sex, even in an
environmental homogenous cohort.
Keywords: skin, canine, microbiota, microbiome, dog, season, skin site, pinna, 16S,
environment
Introduction
Skin is a complex ecosystem inhabited by a high diversity of microorganisms, collectively
referred to as the microbiota. These microbial communities not only inhabit, but also
interact with the host cells impacting cellular function and immunity; likewise the host cells
influence the microbes (1). This cross-talk between the host cells and the microorganisms
maintains the homeostasis and the healthy status of an individual, and when disrupted
usually indicates disease (2).
The dense fur that covers almost all of a dog’s skin creates a homogenous
microenvironment. However, some skin diseases show a preference for certain skin sites
and for specific breeds (3). Previous studies have described skin microbiota on healthy
dogs (4–9) , but only three of them included several skin sites to assess differences that may
exist due to the anatomical location sampled (6,8,9). Results from Rodrigues-Hoffmann
and colleagues showed that haired skin regions presented higher diversity values than
mucosal areas and mucocutaneous junctions (6), and a similar result was reported when
comparing the inner pinna and the perianal region (8). No differences among skin sites
were detected when including only dorsal neck, axilla and abdomen (9).
Dog skin microbiota studies aimed at detecting differences between health and disease
status have already been performed for canine atopic dermatitis (6,7,10). Skin affected with
atopic dermatitis in dogs presented a less diverse microbiota (6,7) and increased
proportions of Staphylococcus and Corynebacterium (7). Moreover, dogs with allergen-induced
atopic dermatitis presented higher proportions of Staphylococcus on the challenged site
compared to the contralateral site (10).
In humans, skin microbiota differs among skin sites and among individuals (11). On one
hand, skin presents three main microhabitats depending on the physicochemical properties:
sebaceous sites, inhabited with Propionibacterium spp; moist sites, with Staphyloccocus and
Corynebacterium spp; and dry sites, with gram-negative microorganisms (11,12). On the other
hand, individual signatures of the skin microbiota are usually driven by low abundant
species (13). Following those first human studies describing skin microbiota, research then
targeted key variables to ascertain if they drove skin microbiota structure and composition
in the healthy individual. Variables assessed and found to have some effect on microbiota
diversity, composition and structure included those related to host such as sex (14–16), age
(17–19) and racial origin (20–22); or related to environment such as birth delivery mode
(23), hygiene (14,22), cohabitation (5,24), geography (21,25,26) and urbanization (19,27,28).
One of the major concerns when performing skin microbiota studies on dogs is the likely
influence of the environment (29). Our previous results suggest that the individual
(together with its environment) was the main force driving skin microbiota composition
and structure in a population of dogs from three different breeds and hair coats (8).
Rodrigues-Hoffmann and colleagues assessed some environmental variables, such as
presence of fleas, time spent indoors vs outdoors, sex, or age and did not detect significant
associations between the microbiota and a particular environmental factor (6). However,
the dog cohort assessed was very variable and included 12 individuals from different
breeds, ages, and households and therefore it is likely that any environmental effects may
have been obscured. Two studies reported that dogs cohabiting together shared more skin
microbiota (5,9). A recent study using a cohort of 40 healthy dogs sampled in three skin
sites across seasons assessed the effects of age, sex, breed, hair type, skin site, season at
time of collection and cohabitation. They found that season was the only variable
significantly stratifying microbiota community structure and samples from different skin
sites were more similar within the same dog (9).
Skin microbiota has been suggested as a potential clinical tool in susceptibility, diagnosis,
and treatment of dermatological diseases (30), therefore characterizing the variability of
skin microbiota in healthy dogs and determining which host and environmental variables
are defining its structure and composition will extend the background to better design
studies aimed to assess the altered skin microbiota on disease.
Here we aimed to assess the variability of the canine skin microbiota in healthy dogs
cohabiting together by analyzing eight different skin sites in a large and homogeneous
cohort of Golden-Labrador Retriever crossbred dogs (N=35). As most of the
environmental variables were fixed, we also aimed to elucidate if any of the host factors
were driving skin microbiota structure and composition in some skin sites. Finally, we
aimed to assess the effect of the geographical region, thus we compared the findings from
the US cohort with those obtained for a group of European dog samples.
Individuals included and sample collection
A cross-sectional study was performed in 35 healthy dogs to analyze skin microbiota
variability in eight skin sites. All dogs were companion dogs, Golden-Labrador Retriever
crosses, with ages ranging from 1 year and 7 months to 2 years and 3 months. They were
born in different households, where they had been raised until 8 weeks of age, and then
they had gone to individual puppy raisers until entering training at ~18 months of age. This
cohort was composed of 20 female and 15 male dogs living and playing together in a
shared environment in Santa Rosa, California. 6 females and 6 males presented a black
coat, and the rest presented a yellow one. This will be named the USA cohort. Additional
File 1 and Additional File 2 contain all the metadata associated with the dogs.
Skin microbiota samples were collected from eight regions taken from the right side of the
dog: inner pinna, chin, nasal skin, back, axilla, abdomen, interdigital area and perianal
region. These regions are named as A, B, C, D, E, F, G and H respectively (Figure 1A).
Samples were obtained by firmly rubbing each area using Sterile Catch-All™ Sample
Collection Swabs (Epicentre Biotechnologies) soaked in sterile SCF-1 solution (50 mM Tris
buffer (pH=8), 1 mM EDTA, and 0.5% Tween-20). To minimize sample cross-
contamination, the person sampling wore a fresh pair of sterile gloves for each individual.
To minimize bias in sampling, only AO and AC sampled the dogs. Swabs were stored at
4ºC until DNA extraction, within the following 8 days (3 days, 2-day stop, 3 days).
On the other hand, the European cohort included 11 dogs of different breeds (Beagle,
French Bulldog, German Shepherd, and West Highland white terrier), ages and
households. Nine of the dogs were previously published and described (8), whereas two of
them sampled later remain unpublished. All samples were processed following the same
DNA extraction protocol, PCR and sequencing approach. All samples were analyzed
together following the steps explained below.
Figure 1. (A) Skin sites sampled and (B) Taxonomic composition per sample included at phylum level.
DNA extraction
Bacterial DNA was extracted from the swabs using the PowerSoil™ DNA isolation kit
(MO BIO) under manufacturer’s conditions, with one modification. At the first lysis step,
the swab tip with the sponge was cut and placed in the bead tube, until the first
transference of the supernatant to a new tube. The remaining steps were performed as
described by the manufacturer in exception of the elution step, which was performed on
50µL of C6 instead of 100µL to obtain a higher concentration. Samples from different skin
sites and individuals were randomly extracted to avoid confounding a batch effect with an
actual effect. DNA extractions were performed within the following 8 days in random
batches of samples to avoid confounding technical biases with actual ones. DNA samples
(50 μl) were stored at -20°C until further processing. To assess for contamination from the
laboratory or reagents two blank samples were processed: one with a sterile swab tip and
the other without the sterile swab tip.
PCR amplification and massive sequencing
V1-V2 regions of 16S rRNA gene were amplified using the widely used primer pair F27 (5’-
AGAGTTTGATCCTGGCTCAG-3’) and R338 (5’-TGCTGCCTCCCGTAGGAGT-3’).
We choose V1–V2 hypervariable regions because they had been suggested to be a better
choice for human skin microbiota among others (31). PCR mixture (25 μl) contained 2 μl
of DNA template, 5 μl of 5x Phusion® High Fidelity Buffer, 2.5 μL of dNTPs (2 mM), 0.2
μM of each primer and 0.5 U of Phusion® Hot Start II Taq Polymerase (Thermo Fisher).
The PCR thermal profile consisted of an initial denaturation of 30 sec at 98 °C, followed by
30 cycles of 15 sec at 98 °C, 15 sec at 55 °C, 20 sec at 72 °C and a final step of 7 min at 72
°C. Samples that did not amplify the first time were repeated increasing cycles to 33. To
assess possible reagent contamination, each PCR reaction included a no template control
(NTC) sample.
For each amplicon, quality and quantity were assessed using Agilent Bioanalyzer 2100 and
QubitTM fluorometer, respectively. Both primers included sequencing adaptors at the 5′
end and forward primers were tagged with different barcodes to pool samples in the same
sequencing reaction.
Each sequencing pool included forty barcoded samples that were sequenced on an Ion
Torrent Personal Genome Machine (PGM) with the Ion 318 Chip Kit v2 and the Ion
PGM™ Sequencing 400 Kit (Life Technologies) under manufacturer’s conditions.
Quality control of the sequences and OTU picking
Raw sequencing reads were demultiplexed and quality-filtered using QIIME 1.9.1 (32).
Reads included presented: a length greater than 300 bp; a mean quality score above 25 in
sliding window of 50 nucleotides; no mismatches on the primer; and default values for
other quality parameters. After that, quality-filtered reads were processed using vsearch
v1.1 pipeline (33): a first de-replication step was applied, followed by clustering into
operational taxonomic units (OTUs) at 97% similarity with a de novo approach and finally
chimera checking was performed using uchime de novo. The raw OTU table was transferred
into QIIME 1.9.1 and taxonomic assignment of representative OTUs was performed using
the Ribosomal Database Project (RDP) Classifier (34) against Greengenes v13.8 database
(35). Alignment of sequences was performed using PyNast (36). We sequentially applied
some extra filtering steps in aligned and taxonomy-assigned OTU table to filter out: 1)
sequences that belonged to Chloroplasts class; 2) sequences representing less than 0.005%
of total OTUs (as previously done in (37)); 3) sequences that belonged to Shewanellaceae and
Halomonadaceae families, which were highly represented in the NTC of the repetition chip
(performed with an increased cycle number) and considered contamination from the
reagents.
Samples 17G and 27A did not amplify and they could not be sequenced. We performed
downstream analysis at a depth of 11,000 sequences per sample: 1D, 30C, 6G and 8G
failed this parameter and were discarded for posterior analyses. Also, NTC and Blank with
a swab tip (S-blank) presented some amplification, but failed to reach 11,000 sequences per
sample; Blank without the swab tip (N-blank) could not amplify.
Downstream bioinformatics analyses
Downstream analyses were performed using QIIME 1.9.1 (32) with the filtered OTU table.
To standardize samples with unequal sequencing depths, analyses were performed using
random subsets of 11,000 sequences per sample.
Alpha diversity analysis assesses the diversity within a sample. Two different metrics were
used for the alpha diversity: observed species to assess richness and Shannon index to
assess evenness. We assessed statistical significant differences in alpha diversity values
among groups with 999 permutations using the non-parametric Monte Carlo permutation
test and corrected the p-value through false discovery rate.
Beta diversity analysis assesses the similarities among samples of the same community. Beta
diversity was performed using both weighted and unweighted UniFrac distance metrics
(38). Weighted UNIFRAC considers phylogeny, taxa and relative abundances; whereas
unweighted UNIFRAC only considers phylogeny and taxa. Those distance matrices were
used to create PCoA plots. ANOSIM and adonis statistical methods were applied to
evaluate if some variables were determining grouping and to which extent.
Linear Discriminant Analysis (LDA) Effect Size (LEfSe) (39) was used to compare groups
and to identify differentially abundance distribution in taxa (α=0.05 and with an LDA score
> 3.0).
CoNet (40), which is implemented as an application in Cytoscape (41), was applied to infer
networks among skin sites using bacterial families that presented a median relative
abundance higher than 0.05% in each specific site. In CoNet we used a combination of five
different algorithms (Pearson’s correlation, Spearman’s correlation, Kullback-Leibler
dissimilarity distances, Bray-Curtis dissimilarity distances and mutual information
similarity). The results of the five methods were merged using Simes p-value. We
performed a first permutation step, followed by a bootstrap analysis corrected for false
discovery rate (α=0.05).
Results
We analyzed the variability of the canine skin microbiota in eight different skin sites from a
healthy homogenous and well-controlled cohort of Golden-Labrador Retriever crossbred
dogs cohabiting together in the same kennel in the United States (N=35) (see Additional
File 1 and Additional File 2 for the associated metadata). We sampled microbiota from
eight skin sites: inner pinna, chin, nasal skin, dorsal back, axilla, abdomen, interdigital
region and perianal area, which are respectively named as A, B, C, D, E, F, G and H
(Figure 1A). These anatomic sites were selected to represent the regional diversity of the
canine skin (3) and to compare with our previous study (8). Samples 17G, 27A, 1D, 30C,
6G and 8G failed at some processing point and were discarded for posterior analyses (see
material and methods for more detail).
We found a total of 2,216 bacterial OTUs living on dog skin (Additional File 3) that were
taxonomically classified into 17 phyla, 41 classes, 62 orders, 128 families and 242 genera.
The abundance of the main phyla differed on each sample (Figure 1B) and were
Proteobacteria (median: 33%; range: 0-99%), Firmicutes (median: 17%; range: 0-97%),
Bacteroidetes (median: 12%, range: 0-74%) Actinobacteria (median: 5%; range: 0-95%),
Cyanobacteria (median: 5%; range: 0%-5%) and Fusobacteria (median: 3%; range: 0-64%),
followed by Tenericutes, TM7 and others with lower abundances.
Alpha diversity values were also very variable among samples with a median value of 5.9
for Shannon index and 414.37 for observed species. The range is broad and goes from 27.8
observed species and 0.27 Shannon index in axilla of Dog 25 to 989.9 observed species in
chin of Dog 33 or 8.5 Shannon index in dorsal back of Dog 34 (Additional File 4A). None
of the dogs was significantly more or less diverse than any other, because there were large
differences in diversity values within the same dog. Those dogs that could seem less diverse
because most of the skin sites presented less diversity, usually presented average values in
inner pinna or perianal region giving no statistical significant differences among individuals
(Additional File 4A).
We performed sample clustering to assess if the variability of the dog skin microbiota
depended on the individual and/or the skin site and to which extent, assessing statistical
significance using Adonis and ANOSIM tests. We also performed an independent analysis
for each skin site to assess the effect of the individual-specific variables, such as sex, coat
color, season of birth or time spent in the kennel.
Individual is driving skin microbiota structure and composition, followed
by skin site
The dogs included in this study were all Golden-Labrador crossbreds with similar ages (~2
years old) and interacted and lived together in a shared environment. Moreover, in most
cases a dog shared genetic background with others: 33 out of 35 dogs presented at least
one half-sibling or littermate and only dogs 31, 19 and 14 were born from different sets of
progenitors (Additional File 2). Grouping the samples per individual significantly explained
23% and 22% of the variation in Unweighted and Weighted UniFrac distance matrices
(Table 1), suggesting that the main force driving the variability of skin microbiota in our
samples was the individual. On the other hand, clustering samples per skin site explained
12% and 17% of the variation respectively and even presented an ANOSIM R value larger
(R=+0.21) than the individual (R=+0.14) when considering OTUs abundance with the
Weighted UNIFRAC distance matrix. Thus, composition of skin microbiota (Unweighted
UniFrac) on healthy dogs was better explained by the individual, whereas structure of the
skin microbiota when abundance was taken into account (Weighted UniFrac) was better
explained by both individual and skin site.
When comparing individual profiles at the phylum level (Figure 1B), we detected that
Proteobacteria was usually the main phylum found on the skin of our cohort: 28 out of 35
dogs presented a higher median value of Proteobacteria than any other phylum. For 5 of
the sampled dogs, Firmicutes was the main phylum. Fusobacteria were most frequently
found in the perianal regions, however when Fusobacteria colonized the haired-skin, the
distribution was individual-specific. That is, there were a few individual dogs with a high
abundance of Fusobacteria in several regions whereas other dogs had almost no
Fusobacteria. Within the Fusobacteria enriched individuals, usually the highest percentages
were found in the abdomen samples. Finally, Cyanobacteria phylum was mainly present
with high abundances in the abdomen, interdigital region and the chin of specific
individuals.
Table 1. Clustering of the samples per biological and technical variables. Beta diversity statistics ANOSIM and Adonis values. (**) p-value=0.001, (*) p-value<0.05 (-) indicates no significant clustering.
Unweighted UniFrac Weighted UniFrac
adonis R2 ANOSIM R adonis R2 ANOSIM R
Individual 0.23** 0.20** 0.22** 0.14**
Skin site 0.12** 0.19** 0.17** 0.21**
Storage time 0.05** 0.07** 0.05** 0.05**
Chip 0.03* 0.03* - -
Person extracting 0.02* - 0.02* -
Sampler 0.01* 0.03* 0.01* 0.03*
Analysis of individual-specific variables
In order to assess if any individual-specific variable defined the skin microbiota
composition or structure in any of the skin sites, we inspected the alpha and beta diversity
of each skin site grouped by the different dog-specific variables such as sex, coat color,
season of birth, time spent in the kennel, or recent surgery and assessed statistical
significance except for the recent surgery due to the small sample size. Depending on the
season of birth, the dogs can be classified in two groups: dogs born from January to May
and dogs born from June to September. The time spent in the kennel coincided with the
season of birth because older dogs (born from January to May) had been in the kennel for
at least 6 months, whereas younger dogs (born from June to September) had been in the
kennel for 3 months. We also considered the pedigree information of the dogs (Additional
File 2): most of the dogs shared some genetic background to at least one other dog (half-
siblings) even across the two main groups.
The season of the year when the dog was born or the time spent in the kennel significantly
affected the microbiota composition (Unweighted UniFrac) and also the community
structure (Weighted UniFrac) in all skin sites (Table 2; Additional File 5). This effect was
especially large on the inner pinna, almost coincident with PC1 component, explaining 26%
of the variation among samples and with an ANOSIM R value of 0.84 (Table 2, Figure
2A). In the other skin sites, these two variables explained more than 9% of the variation
(except for nasal skin), with an R-value ranging from 0.24 to 0.38.
Either the season of birth or the time spent in the kennel was the variable that explained
ubiquitously a significant amount of variation for all the skin sites. Delving into the effect
of these variables on the inner pinna skin microbiota, we corroborated the pattern in the
unweighted UniFrac consensus tree (Figure 2B): two clear clusters were elucidated
matching with the season of birth or the time spent in the kennel (except Dog 8).
Moreover, littermates were usually as similar as any other dog in the same group (except
Dog 2 and 3) and sharing the sire did not make dogs resemble more in skin microbiota.
Moreover, dogs born from January to May or dogs that had spent at least six months in the
kennel were significantly more diverse than the other group (Figure 2C). Finally, LEFSe
analysis detected 61 families differentially distributed in inner pinna when clustering in
these two groups (Additional File 6) and those with higher relative abundances are plotted
in Figure 2D. The greatest difference is provided by Sphingomonadaceae that is highly present
in the inner pinna of individuals born from January to May that had been in the kennel for
6 months, whereas it is almost absent on those dogs born from June to September that had
been in the kennel for 3 months.
Table 2. Individual-specific variables that cluster samples in specific skin sites.
Unweighted UniFrac Weighted UniFrac
Skin site Variable ANOSIM
R adonis R2
ANOSIM R
adonis R2
Inner Pinna Season of birth / Time in the kennel
0.84** 0.26** 0.41** 0.22**
Axilla Season of birth / Time in the kennel
0.38** 0.11** 0.09* 0.07*
Dorsal back
Season of birth / Time in the kennel
0.37** 0.13** 0.28** 0.14**
Interdigital Season of birth / Time in the kennel
0.28** 0.11** 0.09* 0.07*
Abdomen Season of birth / Time in the kennel
0.28** 0.10** 0.09* 0.07*
Perianal Season of birth / Time in the kennel
0.27** 0.09** - -
Chin Season of birth / Time in the kennel
0.24* 0.10* 0.10* 0.08*
Abdomen Sex 0.13* 0.05* 0.24* 0.11**
Nasal skin Season of birth / Time in the kennel
0.11* 0.05* 0.06* -
Back Sex - 0.05* - -
Axilla Sex - - - 0.06*
Figure 2. Season of birth or time spent in the kennel effect on inner pinna. Color blue represents Jan-May group that had been in the kennel for at least 6 months and color red, Jun-Sep group that had been in the kennel for 3 months. (A) Unweighted UniFrac PCoA beta diversity plot. (B) Unweighted UniFrac consensus tree: dogs sharing sire present same-colored branches and littermates are circled and colored with a common pattern within a group (C) alpha diversity rarefaction curves using observed species metrics, and (D) Boxplots of the main differentially distributed families: those include families with abundances > 1% in any
group and also LefSe significant (LDA-score > 3.0, p-value < 0.05).
The sex of the dog also explained some variation. The microbiota community structure in
the abdomen was better explained by the variable sex (11% of the variation in the weighted
UNIFRAC plot and ANOSIM R value +0.24) rather than the season of birth or time spent
in the kennel. This variable also explained to a lesser extent some variability of the
microbiota composition in the dorsal back and the community structure in the axilla (Table
2). Considering the three skin sites affected by sex (abdomen, back and axilla), we could see
that males had an overrepresentation of bacteria from Fusobacteria phylum, with Sneathia
and Fusobacterium genera; also other genera such as Actinomycetospora, Gemella, Parvimonas,
Brevundimonas and phylum SR1 were also overrepresented on males. Females had an
overrepresentation of Enterobacteriaceae family (Table 3).
Table 3. Differentially abundant taxa associated to sex.
Abdomen Axilla Back
Phylum Family or genus Female Male Female Male Female Male
Fusobacteria Fusobacteriales
(order) 1.70% 21.45% 1.64% 13.44% 3.54% 9.55%
Fusobacteria Leptotrichiaceae 0.23% 2.70% 0.24% 3.53% 0.82% 3.00%
Fusobacteria Sneathia 0.01% 0.34% 0.05% 0.25% 0.21% 0.47%
Fusobacteria Fusobacterium NS NS 1.41% 9.91% 2.72% 6.54%
Actinobacteria Actinomycetospora NS NS 0.00% 0.04% 0.00% 0.15%
Firmicutes Gemella 0.19% 3.04% 0.51% 1.61% NS NS
Firmicutes Parvimonas NS NS 0.16% 1.82% 0.55% 1.15%
Proteobacteria Brevundimonas NS NS 0.00% 0.01% 0.00% 0.01%
SR1 SR1 NS NS 0.05% 0.19% 0.14% 0.44%
Proteobacteria Enterobacteriaceae 14.08% 1.31% 7.78% 0.69% NS NS
Relative abundances of main taxa found to be differentially distributed (LDA-score >3, p-value <
0.05) between males and females in at least two out of the three skin sites affected. NS, stands for
no significant differences.
We delved deeper into the five dogs that had undergone surgery, followed by a medical
treatment prior to sampling (Table 4 and Additional File 4B). Dogs 14, 15, 16 and 17
presented reduced alpha diversity values in several skin sites, being chin and abdomen
always affected; whereas alpha diversity values of inner pinna, nasal skin and back were not
reduced in any dog. Dog 20, who underwent surgery three months before sampling,
presented average alpha diversity values.
Finally, the coat color was not significantly explaining the skin microbiota structure or
composition in any skin site.
Table 4. Information of the dogs that had undergone surgery prior to sampling.
Individual Surgery date
Surgery type
Medicines From / To Sites w. reduced α-diversity1
Dog 14 08/04/2016 Spay Amoxicillin (antibiotic) + Previcox (antiinflamatory)
08/04/2016 13/04/2016
Chin and abdomen
Dog 15 18/04/2016 Spay Amoxicillin (antibiotic) + Previcox (antiinflamatory)
18/04/2016 23/04/2016
Chin, axilla, abdomen and ID region
Dog 16 30/03/2016 GI obstruction
Pepcid AC (antihistamine) + Tramadol (analgesic)
01/04/2016 06/04/2016
Chin, axilla, abdomen, ID region and perianal area
Dog 17 12/04/2016 Spay Previcox (antiinflamatory)
12/04/2016 16/04/2016
Chin, axilla, abdomen and perianal
Dog 20 05/01/2016 Spay Rimadyl (antiinflamatory)
05/01/2016 10/01/2016
None
1Reduced alpha diversity values include those ones that are half or less than the median alpha
diversity of that specific skin site of the dogs that had not undergone recent surgery (Additional File
4B).
Influence of other variables on the skin microbiota
The main factors driving skin microbiota in our cohort were the individual and the skin site
as sample associated variables and the sex, the season of birth or the time spent in the
kennel as individual associated variables. However, other sample associated variables
influenced albeit to a lesser extent, with all of them explaining 5% or less of the variation in
the PCoA plots (Table 1). The most significant variable was storage time of the sample
before DNA extraction, with extractions undertaken within the first 3 days being
significantly more diverse than samples extracted in the last 3 days, both in terms of
observed species and Shannon Index (Additional File 4C). Of note, the samples from dogs
that had previously undergone surgery were extracted in the last 3 days, so the analyses
were repeated excluding those individuals and there were still significant differences due to
storage time (Additional File 4D).
We could also detect person extracting was influencing the diversity found in the samples,
that was probably not a real effect though. Differences are detected among AC and AI
extractions: AC was extracting the first 3-days, while AI extracted the last 3-days. So the
samples from that one who extracted the last 3-days are less diverse probably due to
surgery dogs included and due to the later extraction.
Skin sites: Network analysis
A network analysis detects bacterial relationships, within and among different ecological
niches. The global network for all the skin sites considering the most abundant families
allowed us to understand more deeply skin microbiota relationships in our cohort (Figure
3, Table 5 and Additional File S7). Some bacterial species interacted specifically in the same
skin site, whereas other bacterial species interacted among different skin sites. Thus, we
have different ecological niches within the skin.
Chin, abdomen, axilla and perianal region had the highest number of interactions, with 373,
226, 179 and 93 respectively, and also some extra interactions among families of other skin
sites (Table 5 and Additional File 7). On the other hand, inner pinna, nasal skin, interdigital
region and dorsal back presented a lower number of interactions and no inter-site
interactions, as shown in Figure 2. Inner pinna had 35 family interactions; interdigital
region, 23; nasal skin, 7; and dorsal back, only 2.
Table 5. Summary statistics of microbial interactions in the skin of a cohort of healthy dogs.
Chin Abdo-
men Axilla
Perianal region
Inner Pinna
Nasal skin
ID area Dorsal
back
Total Interactions
373 226 179 93 35 7 23 2
Common Interaction
139 103 104 43 13 4 13 2
Unique Interactions
234 123 75 50 22 3 10 0
Inter-site interaction*
3 20 10 12 0 0 0 0
% of unique interactions
63% 54% 42% 54% 63% 43% 43% 0%
% of co-occurence
92% 88% 79% 100% 100% 100% 100% 100%
ID stands for interdigital.*Inter-site interactions represent families from a specific skin site affecting
other families from another skin site
In some cases, specific taxonomic interactions were found within different skin sites. We
identified six interactions highly spread among different skin sites (present in 4 out of 8
skin sites): Neisseriaceae -> Weeksellaceae; Neisseriaceae -> Xenococcaceae (in chin, axilla,
abdomen and perianal); Sphingomonadaceae -> Caulobacteraceae (in inner pinna, axilla,
abdomen and perianal); Sphingomonadaceae -> Nocardioidaceae (in inner pinna, axilla,
abdomen and interdigital region); Sphingomonadaceae -> Oxalobacteraceae (in inner
pinna, chin, abdomen and interdigital); and Weeksellaceae -> Flavobacteriaceae (in chin,
axilla, abdomen and interdigital). However, most interactions (517 out of 703) were
exclusive from one specific skin site (Additional File 7).
This global network demonstrated that most interactions in canine skin were co-ocurrence
relationships rather than mutual exclusion. Among mutual exclusion interactions, few
nodes were negatively linked to many different families within a skin site (circles marked
with a wider black line in Figure 3), showing an apparent invasive pattern. That was seen
for Pseudomonadaceae family in axilla, chin and abdomen and also for Enterobacteriaceae
family in abdomen. When blasting the most abundant OTUs from the highly connected
mutual exclusion nodes, we found that the main genera were Pseudomonas (for
Pseudomonadaceae) and Erwinia and Pantoea (for Enterobacteriaceae) (Additional File 8).
Figure 3. Significant co-occurrence and co-exclusion interactions among the abundant families (>0.005%) in the dog skin microbiota. Nodes are colored depending on the skin site they are found; nodes with a black circle are those highly connected mutual exclusion nodes; edges are green to represent co-occurrence patters and red to represent co-exclusions. Data associated with the complete network can be found at Additional File 7.
Skin sites: taxonomy and diversity analysis
Alpha and beta diversity analyses were undertaken to create an overall characterization of
the canine skin microbiota that accounted for skin sites. Differences in alpha diversity
among skin sites were prevalent. The inner pinna displayed the greatest diversity when
compared to all the other sites and was statistically different to all but the chin site (p-value
= 0.028). The chin, when considering observed species, was significantly more diverse than
the axilla (p-value = 0.028) (Additional File 4C). As seen previously in Table 1, clustering
samples per skin site explained up to 17% of the differences in beta diversity analysis when
considering OTUs abundance (Weighted UniFrac). Differences in microbiota structure
were also significant among almost all skin sites, with the exception of the interdigital
region when compared to abdomen or axilla. We found the greatest differences when
comparing any skin site to the perianal region followed by the nasal skin (Additional File 9).
Focusing on taxonomic analyses, we found that bacteria from Gammaproteobacteria class
were the most abundant in dog skin microbiota, with the exception of perianal regions
where Bacilli class from Firmicutes phylum were the most abundant.
Skin sites shared most of the taxa, but presented also specific taxonomic patterns: the
abundance and distribution varied significantly among skin sites and unique taxa were
identified in some of the sites. Figure 4A shows different bar plots, colored by the main
families found in skin. The families that were differentially distributed (LDA score >3, p-
value = 0.05) are shown in Additional File 10.
The inner pinna had a higher amount of Proteobacteria phylum when compared to other
skin sites, with Gammaproteobacteria, Alphaproteobacteria and Betaproteobacteria classes
being the main representatives. Bacilli (Firmicutes) and Flavobacteriia (Bacteroidetes) were
present in similar abundances to Proteobacteria. Moreover, inner pinna presented many
different and less abundant bacteria.
The chin region was enriched in Gammaproteobacteria, with Pseudomonadaceae as the
main representative family. The nasal skin was also enriched in Gammaproteobacteria, but
the main representative family was Pasteurellaceae. Both families were differentially
distributed in their respective skin site.
Back and axilla had quite similar taxonomic patterns: the main bacterial class was
Gammaproteobacteria, with Moraxellaceae as the main family, followed by Bacilli, with
Lactobacillaceae as one of the main families. The greatest taxonomic difference between both
sites was the higher abundance of Staphylococcaceae (Bacilli class) in the axilla, which was also
differentially distributed when compared to the other skin sites.
The abdomen and interdigital regions had similar taxonomic patterns, where most of the
bacteria were Gammaproteobacteria, specifically from Enterobacteriaceae, Moraxellaceae and
Pseudomonadaceae families; followed by Cyanobacteria, specifically Xenococcaceae family.
However, Planococcaceae was found in abdomen but not in interdigital region.
Finally, the perianal region was the skin site that presented the most differentiated pattern
in dog skin microbiota. The main phylum was Firmicutes, especially Bacilli, followed by
Actinobacteria. Many different families from different phyla were differentially distributed
in the perianal region, indicating that it was the most divergent skin site (Additional File
10). Most of the abundant families in perianal region were also differentially distributed
when compared to the other skin sites. Some of them were: Erysipelotrichaceae,
Lachnospiraceae, Lactobacillaceae and Veillonellaceae (Firmicutes); Corynebacteriaceae
(Actinobacteria); and Bacteroidaceae (Bacteroidetes). The perianal region was also enriched in
Fusobateriaceae, despite not being statistically differentially distributed when compared to
other skin sites.
Figure 4. Taxonomic profiles per skin site. Taxa summary bar plots per class colored by main families within each skin site.
Influence of sample geographical origin: abdomen and dorsal back
To assess if the geographical origin of the samples had an effect on the microbiota, we
performed an additional analysis with only abdomen and back regions, comparing this USA
cohort with some European dogs (see Material and Methods for more details). While the
USA cohort presented uniform characteristics (same crossbred dogs, similar ages, shared
environment, etc.), the European cohort was much more variable including individuals
from different breeds, ages and inhabiting in different households. Thus, we would expect
a clear cluster for the USA controlled cohort and a diffuse if any clustering of Europe
individuals. Samples were processed following the same protocol and the European cohort
was re-analyzed together with the USA cohort, following the previously described analyses
in Material and Methods section.
We visualized two clusters in both abdomen and back when considering the composition
(Unweighted UniFrac), but those clusters lost power when analyzing the community
structure (Weighted UniFrac). The grouping of back samples when considering the
geographical origin was strong (ANOSIM R=+0.68) and it significantly explained 12% of
the variation in Unweighted UniFrac (Additional File 11A). In abdomen samples, grouping
by geographical origin was lower (ANOSIM R=+0.26) and only explained 6% of the
variation (Additional File 11B). Therefore, the back was better at explaining the differences
in microbiota composition among geographical origin than the abdomen site. However, it
is important to note that the back samples composition and structure is more uniform and
less variable among individuals when compared to abdomen samples.
In conclusion, both in back and abdomen samples, the geographical origin explained some
variation that was larger when looking at microbiota composition (Unweighted UniFrac).
Thus, even when comparing two differently constituted cohorts, where European dogs
were not homogenized by any variable, two clusters were clearly identified in beta diversity
analyses representing geographical origin.
Discussion
Our results suggest that the main force driving the skin microbiota composition is the
individual, rather than the skin site, even in a homogeneous cohort of dogs cohabiting and
interacting together. This is in line with what we found previously in a cohort of nine
healthy dogs from three different breeds, although in that study we could not elucidate
whether the individual effect was real or represented an environmental influence (8). Here,
we homogenized the cohort to account for different effects: same crossbreed dogs, same
age and environment. Our results confirm the individual as the main factor determining
skin microbiota composition in healthy dogs, when abundance of the bacterial species was
not taken into account (Unweighted UniFrac). In human skin microbiota, low abundant
species are those defining individuality and their signatures have the ability to identify items
that the person came in contact with (13,42,43) and are relatively constant over time (44).
An individual effect had also been reported as the main driver of fungal skin microbiota
structure and composition in dogs from heterogeneous cohorts (45) and had been
suggested to affect also bacterial skin microbiota in dogs, despite the individual was not
assessed directly (6).
Our results also suggest that when relative abundances of the bacteria were taken into
account (Weighted UniFrac), both skin site and individual affected the skin microbiota
structure of healthy dogs. Similarly, these two factors also shaped human skin microbiota,
with great variability within several skin sites of an individual and between individuals
having been reported (11,46,47). Human skin has three main microhabitats (moist, dry and
sebaceous) inhabited by specific taxa (12,48). Although the three microhabitats clearly
identified in humans were not seen in dogs (29), Rodrigues-Hoffmann and colleagues
reported significant differences between haired and mucosal or muco-cutaneous junctions
(6), which coincides with our current observation. Here, we found that perianal region and,
to lower extent, nasal skin presented different community structure (Weighted UniFrac) as
well as lower alpha diversity values when compared to all other haired skin regions.
Globally in our cohort, Gammaproteobacteria followed by Bacilli were the most abundant
classes in all regions in exception of perianal region with the same classes but the opposite
order. A previous study including dorsal neck, abdomen and axilla samples from 40
domestic dogs inhabiting different households found Gammaproteobacteria and Bacilli as
main classes, but also Actinobacteria (9). On the other hand, Hoffmann and colleagues (6)
detected different abundant classes depending on the skin site: Betaproteobacteria was the
most common in the concave pinna, dorsal lumbar and ear; Actinobacteria, in the axilla
and interdigital skin; Gammaproteobacteria, in the nostril; and Clostridia and Bacteroidia,
in the perianal region. Finally, in our previous study, we found Bacilli as the main class for
all the skin sites with the exception of inner pinna that had Alphaproteobacteria (8). Thus,
as the inter-individual variability is large, independent studies led to similar results only
when a large number of individuals are included.
Network analysis elucidated the overall community organization throughout the skin of our
canine cohort, with more than 40% of the interactions exclusive of each site,
demonstrating a skin site signature. Back skin presented only two interactions and both of
them were back-exclusive, probably other interactions remain hidden because only
abundant families were included for network analysis. Among all skin sites, the inner pinna
and chin were the sites that presented a higher proportion of unique interactions,
suggesting stronger specialization or influences. On one hand, the inner pinna is an
anatomically and environmentally isolated skin site when compared to others. On the other
hand, we suggest that chin presented influences of both drinking water and oral
microbiota. The most abundant families were Xenococcaceae and Pseudomonadaceae, which had
been isolated in several water sources (49)(50). Moreover, the following abundant families,
such as Fusobacteriaceae, Moraxellaceae, Porphyromonadaceae, Neisseriaceae and Flavobacteriaceae,
were previously found as main taxa in canine oral microbiota (7,51).
Network analysis detected that most of the interactions among abundant families on the
skin of Golden-Labrador Retriever crossbreds were co-occurrence rather than co-
exclusion, which contrasts with what was previously seen for human microbiome with a
balanced ratio of co-occurrence vs co-exclusion relationships within a body site (52).
Despite the majority of co-occurrence interactions, few families presented a high number
of mutual exclusions: Pseudomonadaceae and Enterobacteriaceae in abdomen and
Pseudomonadaceae in axilla and chin. When blasting the OTUs that presented this apparently
invasive pattern (Additional File 8), we found that the ones belonging to
Enterobacteriaceae family had been mainly isolated from soil or plant surfaces (53,54);
whereas those from Pseudomonadaceae family had been mainly isolated from soil and
different sources of water (50). Thus, we suggest that this pattern is representing a recent
exposure to the environment prior to sampling of some dogs. Other bacteria were
suspected to have an environmental origin, despite not showing a co-exclusion pattern,
such as Xenococcaceae with Chroococcidiopsis as its main genus. Bacteria from this genus had
been mainly isolated from freshwater environments including lakes, soil, or inside of rocks
(49). Moreover, they have already been detected on healthy dog skin (6,8). The presence of
these bacteria with high abundance on the interdigital and abdomen regions may suggest
these two regions are more susceptible to environmental influences, which seems
reasonable since these two skin sites have direct contact with the ground.
The skin sites could be classified based upon two patterns. The first pattern included sites
having a high number of interactions among abundant families, with some interactions
with other skin sites (chin, axilla, abdomen, and perianal region). The second pattern
included sites having a lower number of interactions among abundant families and
displayed exclusively within-site interactions (pinna, nasal skin, dorsal back and interdigital
area). We suggest that the inter-site relationships could be explained due to topographical,
behavioral and environmental factors. The chin is juxtaposed to the mouth, which is a
main entrance for the environment through licking, eating, or drinking water. Dogs could
lap the same water in which they are playing, and they usually lick themselves, which could
explain some interactions among those sites. Additionally, the abdomen and axilla are
anatomically continuous on the ventral side of the dog and close to the ground facilitating
interactions with the environment and between the two skin sites. Furthermore, dogs may
come into contact with fecal matter, which could explain shared OTUs among the
abdomen, axilla and perianal regions. Main families of the second pattern, constituted by
the inner pinna, dorsal back, interdigital area and nasal skin, were only interacting with
other families in the same skin site, suggesting that both anatomical isolation and stronger
effects of other microbiota (nostril microbiota, for nasal skin and soil microbiota for
interdigital region) may account for the exclusive within-site interactions.
With this general overview, we sought to elucidate if any individual-specific variable
determined the observed diversity, composition, and/or community structure in any of the
skin sites. When considering the season of birth or the time spent in the kennel as variable,
we observed two significantly different groups: dogs born from January to May that had
spent 6 months in the kennel were different from those born from June to September that
had spent 3 months in the kennel. This effect was highest on the inner pinna, with a
significant ANOSIM R value of +0.84. Main taxonomic difference among inner pinna
from both groups was due to Sphingomonadaceae, specifically Sphingomonas. These taxa are
classically considered air- and dust-borne (55), although they had also been identified on
dog skin microbiota (5,8) and in animal sheds (55,56), even specifically on dogs’ (57). These
classically considered air- and dust-borne bacteria are cultivable at temperatures ranging
from 4–28 °C, but not at 37 °C (55). Independent studies of grapevine microbiome show a
link between Sphingomonadaceae and lower temperatures: in leaves Sphingobacteriaceae was
significantly overrepresented on samples from May when compared to July (58); and in
grapes, samples from a colder year presented increased levels of Sphingomonadaceae (59).
Thus, we suggest two alternate hypotheses. If the variable explaining those differences was
the season of birth, maybe the maximum temperatures reached in the summer time, usually
overpassing 28ºC, implied lower Sphingomonadaceae presence in the air, in dust and on dogs’
hair coats, avoiding its establishment as a component of the dogs’ skin microbiota. In
human studies, infant skin microbiota resembled the maternal one up to six weeks after
birth, contrasting with microbiota from stool, nares or oral cavity which significantly
differed (60). Thus, applying few concepts of ecology theory (61), we suggest that the
bacterial pool of the environment and the air was shaped with the season characteristics
(humidity, UV light, temperature, etc.), which in turn was shaping to some extent the skin
microbiota of dogs (environmental selection). Thus, when a dog was born it had a different
available bacterial pool depending on the season that will affect the invasion order: some
species were more abundant, so more likely to be the first to colonize the skin. Or perhaps
the dam had an increased number of season-specific bacteria on her shed, transferring
them to the puppies and therefore becoming resident inhabitants of the puppy skin
microbiota by historical contingency processes, where the order of invasion matters and it
could randomly differ among littermates (for example, being born first vs second). If the
time spent in the kennel was the variable explaining these divergences, we could be
detecting again an environmental effect: those dogs that were during the autumn and
winter in the kennel were exposed for a longer period of time to the environment and were
more likely to harbor bacteria from it. In this case we should accept a high influence of the
environment in the skin microbiota composition, entitling even a replacement of the main
families depending on the time of cohabitation and the household placement conditions.
Finally, we should not forget the possible role of progenitors’ genetics, but in both groups
there are some common sires despite different dams discarding the possible effect of sire
genetics. Moreover, inner pinna skin microbiota is not more similar when comparing two
littermate’s than when comparing any other two dogs within the same group (in exception
of Dog 2 and 3).
Although it is difficult to elucidate which bacteria are really microbiota and which others
are only transient members from the environment, the pattern seems clear for
Sphingomonadaceae. All the samples were obtained the same day and a significant difference
was found regarding the season of birth or time spent in the kennel. Therefore,
Sphingomonadaceae, classically regarded as air- and dust-borne bacteria, should be considered
also a normal colonizer of dog skin microbiota. Similarly as dog skin with Sphingomonadaceae,
researchers had found taxa classically regarded as environmental being part of human skin
microbiota. For example, the genus Enhydrobacter was commonly found in air and surfaces
of built environment of Hong Kong (62) and also presented high abundances in skin of
Chinese individuals (21,63). Another example would be Amerindian individuals, which
presented a very different skin microbiota profile with high proportion of bacteria
commonly regarded as environmental (25).
Besides the season of birth or the time spent in the kennel, sex had a significant effect on
the abdomen, back and axilla microbiota of our cohorts. Female dogs presented an
overrepresentation of Enterobacteriales and Enterobacteriaceae families, coinciding with what
was previously reported on the hands of humans (14).
Dogs that had undergone surgery within the previous month presented lower alpha
diversity values, always in abdomen and chin. The surgery procedures that had undergone
implied shaving the abdomen and were followed by oral medication administration
(sometimes antibiotics), which would probably explain the lowered alpha diversity values.
Finally, geography also affected dog skin microbiota, most significantly its composition.
The European cohort was comprised of pet dogs that did not interact with each other,
were from different households, age groups and genetic backgrounds, and their samples
were collected in different seasons of the year. Even considering this heterogeneity, the
European dogs clustered together in a single group differing from the environmental well-
controlled USA cohort. That is in line with what has already been described for humans
with geography (21,26), geographical isolation (25) or urbanization (27,28)(19) grouping
differently skin microbial communities.
Focusing on the technique, our results showed that skin microbiota samples collected with
swabs and stored at 4ºC for more than 5 days presented significantly less diversity when
compared to those extracted sooner; this was true even when the dogs who had surgery,
and were therefore less diverse, were excluded (Additional File 4D). Each day a different
set of random samples was extracted and our results demonstrate that there were no
impacts of extraction within the first 3-days; neither were differences detectable among the
last 3-days. However, microbiota diversity values differed when extraction for weeks one
and two were compared. This contrasts with the results reported by Lauber and colleagues
that assessed stability of the skin microbiota samples when stored in different temperatures.
They found no differences on diversity when they were stored either at 4ºC or 20ºC for up
to two weeks (64). However, they worked at a sequencing depth 10 times lower than our
study: 1,000 vs 11,000 sequences per sample. The lower coverage may have obscured the
loss of diversity because the lower prevalent species may not have been detected at the
lower sequencing depth. Thus, we recommend extracting microbial DNA from skin swab
samples within 3-days of sample collection with storage at 4ºC. Finally, it is important to
note that this decrease in the diversity only slightly affected skin community structure and
composition (explaining ≤5% of the variation).
Conclusions
In summary, we have characterized the normal variability of dog skin microbiota in a well-
controlled cohort of a large number of dogs with similar ages, related genetic background,
and a shared environment. We found that microbiota composition was driven by the
individual, but when considering the abundances, microbiota structure was driven both by
the individual and by the skin site. Network analyses elucidated that both exclusive and
shared interactions existed depending on the skin site, with the highly shared interactions
probably representing an environmental origin. When analyzing each skin site
independently to assess host-specific factors we found that season of birth or time spent in
the kennel affected all skin sites. The most abundant taxon driving this difference was
Sphingomonas, which is an air-borne bacterium that cannot be cultivated at elevated
temperatures. We also found taxonomic differences among male and female dogs on
abdomen, axilla and back. Moreover, the USA and European cohorts were grouping by
geographical origin in two different and well-defined clusters, even when the European
individuals were very heterogeneous. In conclusion, we observed a large inter-individual
variability and effects of different host variables, even in an environmental well-controlled
cohort.
Thus, to overcome the individual variability inherent to skin microbiota studies, we would
recommend longitudinal studies assessing divergences between health and disease
comparing affected vs unaffected regions within an individual through time; or alternatively
the cohort should be large enough and well controlled if case-control studies are preferred.
Understanding the skin microbiota of healthy skin will allow a better knowledge of the
intrinsic variability in health and the assessment of what is an altered state. It will also
provide a background to develop its clinical applications (30) such as identifying an altered
skin microbiota landscape or developing personalized therapies aimed at shifting the
balance toward a healthy skin microbiota, promoting beneficial bacteria growth rather than
killing all bacteria.
Declarations
Ethics approval and consent to participate
All animal work was done with the approval of the University of California, Davis
Institutional Animal Care and Use Committee and the scientific research oversight
committee of Canine Companions for Independence.
The European dogs included in this study were examined during routine veterinary
procedures by the veterinary clinics participating in the study. All samples were collected
and used in the study with verbal owner consent. As the data are from client-owned dogs
that underwent normal preventative veterinary examinations, there was no “animal
experiment” according to the legal definitions in Spain, and approval by an ethical
committee was not necessary.
Consent for publication
Not applicable.
Availability of data and material
The datasets analyzed during the current study are available in the SRA NCBI repository
under the Bioproject accession number PRJNA384381.
Competing interests
The authors declare that they have no competing interests.
Funding
This work was partly supported by a grant awarded by Generalitat de Catalunya (Industrial
Doctorate program, 2013 DI 011).
Authors' contributions
AO, JM, OF, AS and AC conceived and designed the experiment. OF, AO, AS and JM
supervised the project and gave conceptual advice. AO, JB, AI, KL, and AC participated in
the sample collection. JB, AI, LG, and AC performed the DNA extractions. AC performed
the PCRs. KL recollected all the metadata. AC carried out the bioinformatics analysis. AC
drafted the manuscript. OF, AO, JB, and LG edited the manuscript. All authors read and
approved the final manuscript.
Acknowledgements
We would like to acknowledge Canine Companions for the Independence organization for
providing the skin samples from their dogs. We would also acknowledge Nicolas Boulanger
and Joana Ribes for the sequencing service and Gonzalo Vera for the informatics support.
Additional files
They are available at the following link:
https://www.dropbox.com/sh/kcd1bo4adzlh439/AADsgFQ5MwNYTieimIITZUUAa?dl=0
Additional File 1.xlsx - Information about the dogs included in the study. Sample were
collected the 2016/04/27.
Additional File 2.docx - Pedigree chart of the dogs included in this study. Circle represent
female and rectangle represent male. In blue, dogs born from January to May that had
spent at least 6 months in the kennel (Jan-May group) and in red dogs born from June to
September that had spent 3 months in the kennel (Jun-Sep group).
Additional File 3.xlsx - OTU table at genus level including all the samples.
Additional File 4.xlsx - Alpha diversity values and statistics.
Additional File 5.docx - Unweighted UniFrac beta diversity PCoA plot per skin site
colored by season of birth or time spent in the kennel. In red, dogs born from January to
May that had spent at least 6 months in the kennel (Jan-May group) and in blue dogs born
from June to September that had spent 3 months in the kennel (Jun-Sep group).
Additional File 6.docx - Differentially distributed families based on season of birth or
time spent in the kennel. Histogram of linear discriminant analysis (LDA) effect size
(LefSe) for differentially abundance distribution (α = 0.05, LDA score >3).
Additional File 7.xlsx - Network output. CoNet output tables with edge and node
information.
Additional File 8.xlsx - Taxonomies of the highly-connected nodes obtained through
BLAST.
Additional File 9.xlsx - ANOSIM R values for each pair of skin sites and both for
Weighted and Unweighted UniFrac matrices.
Additional File 10.docx - Differentially distributed families based on skin site. Histogram
of linear discriminant analysis (LDA) effect size (LefSe) up to family level for differentially
abundant distributed taxa (α = 0.05, LDA score >3).
Additional File 11.docx - Geographical origin effect on beta diversity for back and
abdomen samples. Samples from this study (USA) were merged with previous samples
(Spain) (Cuscó et al., 2017) as well as two other unpublished individuals. Unweighted
UniFrac beta diversity plots of (A) dorsal back and (B) abdomen samples colored by
geographical origin with their associated ANOSIM and adonis values.
Bibliography
1. Naik S, Bouladoux N, Wilhelm C, Molloy MJ, Salcedo R, Kastenmuller W, et al. Compartmentalized
Control of Skin Immunity by Resident Commensals. Science 2012, 337:1115–9.
2. Cho I, Blaser MJ. The human microbiome: at the interface of health and disease. Nat Rev Genet 2012,
13:260–70.
3. Miller WH, Griffin CE, Campbell KL. Muller & Kirk’s Small Animal Dermatology. 7th ed. St Louis:
Elsevier; 2013.
4. Weese JS. The canine and feline skin microbiome in health and disease. Vet Dermatol 2013, 24:137–46.
5. Song SJ, Lauber C, Costello EK, Lozupone CA, Humphrey G, Berg-Lyons D, et al. Cohabiting family
members share microbiota with one another and with their dogs. Elife 2013, 2:e00458.
6. Rodrigues Hoffmann A, Patterson AP, Diesel A, Lawhon SD, Ly HJ, Elkins Stephenson C, et al. The skin
microbiome in healthy and allergic dogs. PLoS One 2014, 9:e83197.
7. Bradley CW, Morris DO, Rankin SC, Cain CL, Misic AM, Houser T, et al. Longitudinal evaluation of the
skin microbiome and association with microenvironment and treatment in canine atopic dermatitis. J Invest
Dermatol 2016, 136:1182–90.
8. Cuscó A, Sánchez A, Altet L, Ferrer L, Francino O. Individual Signatures Define Canine Skin Microbiota
Composition and Variability. Front Vet Sci 2017, 4:6.
9. Torres S, Clayton JB, Danzeisen JL, Ward T, Huang H, Knights D, et al. Diverse bacterial communities
exist on canine skin and are impacted by cohabitation and time. PeerJ 2017, 5:e3075.
10. Pierezan F, Olivry T, Paps JS, Lawhon SD, Wu J, Steiner JM, et al. The skin microbiome in allergen-
induced canine atopic dermatitis. Vet Dermatol 2016, 5:332–e82
11. Grice EA, Segre JA. The skin microbiome. Nat Rev Microbiol 2011, 9:244–53.
12. Grice EA, Kong HH, Conlan S, Deming CB, Davis J, Young AC, et al. Topographical and temporal
diversity of the human skin microbiome. Science 2009, 324:1190–2.
13. Fierer N, Lauber CL, Zhou N, McDonald D, Costello EK, Knight R. Forensic identification using skin
bacterial communities. Proc Natl Acad Sci U S A 2010, 107:6477–81.
14. Fierer N, Hamady M, Lauber CL, Knight R. The influence of sex, handedness, and washing on the
diversity of hand surface bacteria. Proc Natl Acad Sci U S A 2008, 105:17994–9.
15. Zeeuwen PL, Boekhorst J, van den Bogaard EH, de Koning HD, van de Kerkhof PM, Saulnier DM, et al.
Microbiome dynamics of human epidermis following skin barrier disruption. Genome Biol 2012, 13:R101.
16. Callewaert C, Kerckhof F-M, Granitsiotis MS, Van Gele M, Van de Wiele T, Boon N. Characterization of
Staphylococcus and Corynebacterium Clusters in the Human Axillary Region. PLoS ONE 2013, 8:e70538.
17. Capone KA, Dowd SE, Stamatas GN, Nikolovski J. Diversity of the human skin microbiome early in life.
J Invest Dermatol 2011, 131:2026–32.
18. Oh J, Conlan S, Polley E, Segre JA, Kong HH. Shifts in human skin and nares microbiota of healthy
children and adults. Genome Med 2012, 4:77.
19. Lehtimäki J, Karkman A, Laatikainen T, Paalanen L, von Hertzen L, Haahtela T, et al. Patterns in the skin
microbiota differ in children and teenagers between rural and urban environments. Sci Rep 2017, 7:45651.
20. The Human Microbiome Project Consortium. Structure, function and diversity of the healthy human
microbiome. Nature 2012, 486:207–14.
21. Leung MHY, Wilkins D, Lee PKH. Insights into the pan-microbiome: skin microbial communities of
Chinese individuals differ from other racial groups. Sci Rep 2015, 5:11845.
22. Perez Perez GI, Gao Z, Jourdain R, Ramirez J, Gany F, Clavaud C, et al. Body Site Is a More
Determinant Factor than Human Population Diversity in the Healthy Skin Microbiome. PLoS One 2016,
11:e0151990.
23. Dominguez-bello MG, Costello EK, Contreras M, Magris M, Hidalgo G. Delivery mode shapes the
acquisition and structure of the initial microbiota across multiple body habitats in newborns. Proc Natl Acad
Sci U S A 2010, 107:11971–5.
24. Misic AM, Davis MF, Tyldsley AS, Hodkinson BP, Tolomeo P, Hu B, et al. The shared microbiota of
humans and companion animals as evaluated from Staphylococcus carriage sites. Microbiome 2015, 3:2.
25. Clemente JC, Pehrsson EC, Blaser MJ, Sandhu K, Gao Z, Wang B, et al. The microbiome of uncontacted
Amerindians. Sci Adv 2015, 1:e1500183.
26. Hospodsky D, Pickering AJ, Julian TR, Miller D, Gorthala S, Boehm AB, et al. Hand bacterial
communities vary across two different human populations. Microbiology 2014, 160:1144–52.
27. Hanski I, von Hertzen L, Fyhrquist N, Koskinen K, Torppa K, Laatikainen T, et al. Environmental
biodiversity, human microbiota, and allergy are interrelated. Proc Natl Acad Sci 2012, 109:8334–39.
28. Ying S, Zeng D, Chi L, Tan Y, Galzote C, Cardona C. The Influence of Age and Gender on Skin-
Associated Microbial Communities in Urban and Rural Human Populations. PLoS One 2015, 10:e0141842.
29. Rodrigues Hoffmann A. The cutaneous ecosystem: the roles of the skin microbiome in health and its
association with inflammatory skin conditions in humans and animals. Vet Dermatol 2017, 1:60–e15.
30. Grice EA. The skin microbiome: potential for novel diagnostic and therapeutic approaches to cutaneous
disease. Semin Cutan Med Surg 2014, 33:98–103.
31. Kuczynski J, Lauber CL, Walters WA, Parfrey LW, Clemente JC, Gevers D, et al. Experimental and
analytical tools for studying the human microbiome. Nat Rev Genet 2011, 13:47–58.
32. Caporaso JG, Kuczynski J, Stombaugh J, Bittinger K, Bushman FD, Costello EK, et al. QIIME allows
analysis of high-throughput community sequencing data. Nat Methods 2010, 7:335–6.
33. Rognes T, Flouri T, Nichols B, Quince C, Mahé F. VSEARCH : a versatile open source tool for
metagenomics. PeerJ 2016, 4:e2584.
34. Wang Q, Garrity GM, Tiedje JM, Cole JR. Naive Bayesian classifier for rapid assignment of rRNA
sequences into the new bacterial taxonomy. Appl Environ Microbiol 2007, 73:5261–7.
35. DeSantis TZ, Hugenholtz P, Larsen N, Rojas M, Brodie EL, Keller K, et al. Greengenes, a chimera-
checked 16S rRNA gene database and workbench compatible with ARB. Appl Environ Microbiol 2006,
72:5069–72.
36. Caporaso JG, Bittinger K, Bushman FD, Desantis TZ, Andersen GL, Knight R. PyNAST: A flexible tool
for aligning sequences to a template alignment. Bioinformatics 2010, 26:266–7.
37. Navas-molina JA, Peralta-sánchez JM, González A, McMurdie PJ, Vázquez-Baeza Y, Xu Z. Advancing
Our Understanding of the Human Microbiome Using QIIME. Methods Enzymol 2013, 531:371–444.
38. Lozupone C, Knight R. UniFrac : a New Phylogenetic Method for Comparing Microbial Communities.
Appl Environ Microbiol 2005, 71:8228–35.
39. Segata N, Izard J, Waldron L, Gevers D, Miropolsky L, Garrett WS, et al. Metagenomic biomarker
discovery and explanation. Genome Biol 2011, 12:R60.
40. Faust K, Raes J. CoNet app : inference of biological association networks using Cytoscape.
F1000Research 2016, 5:1519.
41. Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, Ramage D, et al. Cytoscape: a software
environment for integrated models of biomolecular interaction networks. Genome Res 2003, 13:2498–504.
42. Meadow JF, Altrichter AE, Green JL. Mobile phones carry the personal microbiome of their owners.
PeerJ 2014, 2:e447.
43. Lax S, Hampton-Marcell JT, Gibbons SM, Colares GB, Smith D, Eisen JA, et al. Forensic analysis of the
microbiome of phones and shoes. Microbiome 2015, 3:21.
44. Oh J, Byrd AL, Park M, Kong HH, Segre JA. Temporal Stability of the Human Skin Microbiome. Cell
2016, 165:854–66.
45. Meason-smith C, Diesel A, Patterson AP, Older CE, Mansell JM, Suchodolski JS, et al. What is living on
your dog’s skin ? Characterization of the canine cutaneous mycobiota and fungal dysbiosis in canine allergic
dermatitis. FEMS Microbiol Ecol 2015, 91:1–12.
46. Oh J, Byrd AL, Deming C, Conlan S, Barnabas B, Blakesley R, et al. Biogeography and individuality shape
function in the human skin metagenome. Nature 2014, 514:59–64.
47. Ursell LK, Clemente JC, Rideout JR, Gevers D, Caporaso JG, Knight R: The interpersonal and
intrapersonal diversity of human-associated microbiota in key body sites. J Allergy Clin Immunol 2012,
129:1204-8
48. Costello EK, Lauber CL, Hamady M, Fierer N, Gordon JI, Knight R. Bacterial community variation in
human body habitats across space and time. Science 2009, 326:1694–7.
49. Cumbers J, Rothschild LJ. Salt tolerance and polyphyly in the cyanobacterium Chroococcidiopsis
(Pleurocapsales). J Phycol 2014, 50:472–82.
50. Peix A, Ramírez-Bahena M-H, Velázquez E. Historical evolution and current status of the taxonomy of
genus Pseudomonas. Infect Genet Evol 2009, 9:1132–47.
51. Sturgeon A, Stull JW, Costa MC, Weese JS. Metagenomic analysis of the canine oral cavity as revealed by
high-throughput pyrosequencing of the 16S rRNA gene. Vet Microbiol 2013, 162:891–8.
52. Faust K, Sathirapongsasuti JF, Izard J, Segata N, Gevers D, Raes J, et al. Microbial co-occurrence
relationships in the human microbiome. PLoS Comput Biol 2012, 8:e1002606.
53. Palmer M, Steenkamp ET, Coetzee MPA, Chan W-Y, van Zyl E, De Maayer P, et al. Phylogenomic
resolution of the bacterial genus Pantoea and its relationship with Erwinia and Tatumella. Antonie Van
Leeuwenhoek 2017, 1–23.
54. Walterson AM, Stavrinides J. Pantoea: insights into a highly versatile and diverse genus within the
Enterobacteriaceae. FEMS Microbiol Rev 2015, 39:968–84.
55. Busse H-J, Denner EBM, Buczolits S, Salkinoja-Salonen M, Bennasar A, Kämpfer P. Sphingomonas
aurantiaca sp. nov., Sphingomonas aerolata sp. nov. and Sphingomonas faeni sp. nov., air- and dustborne and
Antarctic, orange-pigmented, psychrotolerant bacteria, and emended description of the genus Sphingomonas.
Int J Syst Evol Microbiol 2003, 53:1253–60.
56. Andersson AM, Weiss N, Rainey F, Salkinoja-Salonen MS. Dust-borne bacteria in animal sheds, schools
and children’s day care centres. J Appl Microbiol 1999, 86:622–34.
57. Miletto M, Lindow SE. Relative and contextual contribution of different sources to the composition and
abundance of indoor air bacteria in residences. Microbiome 2015, 3:61.
58. Pinto C, Pinho D, Sousa S, Pinheiro M, Egas C, Gomes AC. Unravelling the Diversity of Grapevine
Microbiome. PLoS One 2014, 9:e85622.
59. Zarraonaindia I, Owens SM, Weisenhorn P, West K, Hampton-Marcell J, Lax S, et al. The soil
microbiome influences grapevine-associated microbiota. MBio 2015, 6: e02527-14.
60. Chu DM, Ma J, Prince AL, Antony KM, Seferovic MD, Aagaard KM. Maturation of the infant
microbiome community structure and function across multiple body sites and in relation to mode of delivery.
Nat Med 2017, 23:314–26.
61. Costello EK, Stagaman K, Dethlefsen L, Bohannan BJM, Relman DA. The application of ecological
theory toward an understanding of the human microbiome. Science 2012, 336:1255–62.
62. Leung MHY, Lee PKH. The roles of the outdoors and occupants in contributing to a potential pan-
microbiome of the built environment: a review. Microbiome 2016, 4:21.
63. Ling Z, Liu X, Luo Y, Yuan L, Nelson KE, Wang Y, et al. Pyrosequencing analysis of the human
microbiota of healthy Chinese undergraduates. BMC Genomics 2013, 14:390.
64. Lauber CL, Zhou N, Gordon JI, Knight R, Fierer N. Effect of storage conditions on the assessment of
bacterial community structure in soil and human-associated samples. FEMS Microbiol Lett 2010, 307:80–6.
111
3.4. Using MinION™ to characterize dog skin
microbiota through full-length 16S rRNA
gene sequencing approach
This chapter consists of the article entitled “Using MinION™ to characterize dog skin
microbiota through full-length 16S rRNA gene sequencing approach” pending to be
submitted.
Using MinION™ to characterize dog skin microbiota through full-length 16S rRNA gene
sequencing approach
Introduction Bacteria, fungi, viruses and archaea are the main microorganisms constituting the microbiota, which is defined as the microbial communities inhabiting a specific environment (1). In humans, many efforts have been made to characterize the different body site ecosystems and their associated microbial communities, mainly at bacterial level (2,3), which are the most abundant microorganisms on the human-associated microbiota (4,5).
Studying host-associated microbiota has provided many insights on health and diseases for many different body sites (6,7). In human skin, alterations on skin microbiota have been associated to numerous cutaneous diseases, such as acne vulgaris (8,9), psoriasis (10–12), or atopic dermatitis (13–17). Not only humans, but also dogs presented altered microbiota states during atopic dermatitis disease (18–20).
The most common strategy to assess bacterial microbiota is amplifying and sequencing specific regions of 16S rRNA gene using 2nd generation massive sequencing technologies (for a review see (21)). This bacterial marker gene is ubiquitously found in bacteria, and has nine hypervariable regions (V1-V9) that can be used to infer taxonomy (22).
The ability to classify sequences to the genus or species level is a function of read length, sample type and the reference database (23). High-quality short-reads obtained from 2nd generation sequencers (250-350 bp) bias and limit the taxonomic resolution of this gene. The more usual regions amplified with Illumina MiSeq or Ion Torrent PGM™ for bacterial taxonomic classification are V4 or V4-V5, but these regions fail in amplifying some significant species for skin microbiota studies, such as Propionibacterium acnes. So, when performing a skin microbiota study the preferred choice is V1-V2 regions, although they lack sensitivity for the genus Bifidobacterium and poorly amplify the phylum Verrucomicrobia (21). On the other hand, near full-length 16S rRNA gene sequences are required for accurate richness estimations especially at higher taxa (24), which are necessary on microbiota studies. Besides, full-length reference sequences are needed for performing phylogenetic analyses or designing lineage specific primers (23), especially in species different to human or mouse, in which previous metagenomics approaches deciphered the richness of bacterial species in the great and different variety of microbiome samples analyzed.
With the launching of 3rd generation single-molecule technology sequencers, these short-length associated problems can be overcome by sequencing the full or almost full-length of 16S rRNA gene with different sets of universal primers (25). Results for full-length 16S rRNA gene have been reported for Pacific Biosciences (PacBio) platform (23,26–30). Schloss and collaborators reported the possibility of generating near full-length 16S rRNA gene sequences with error rates slightly higher, but comparable to the 2nd generation platforms (0.03%)(23). The primary limitation on the PacBio platform is the accessibility to the sequencers and the cost of generating the data.
MinIONTM sequencer of Oxford Nanopore Technologies (ONT) (https://nanoporetech.com) is a 3rd generation sequencer that is portable, affordable with a small budget and offers long-read output (only limited by DNA extraction protocol). Besides, it can provide a rapid real-time and on-demand analysis very useful on clinical applications. Several studies targeting the full 16S rRNA gene have already been performed using MinIONTM to: i) identify pure bacterial culture (31); ii) characterize artificial and already-characterized bacterial communities (mock community) (32–34); and to iii) characterize complex microbiota samples, from mouse gut (35), wastewater (31) and pleural effusion from a patient with empyema (34).
Here we aim to assess the potential of Nanopore sequencing in complex microbiota samples using the full-length 16S rRNA (1,500bp). First set-up step is performed using a staggered mock community (HM-783D). Then, we sequenced a pool of several skin microbiota samples previously sequenced by Ion Torrent PGM™.
Material and methods
Samples included
As simple microbial community, we used a Microbial Mock Community HM-783D kindly donated by BEI resources (http://www.beiresources.org) that contained genomic DNA from 20 bacterial strains with staggered ribosomal RNA operon counts (1,000 to 1,000,000 copies per organism per μL). This mock community allowed us to perform the MinIONTM sequencing and analysis protocol set-up.
As complex microbial community, we used a sample pool from skin microbiota of healthy dogs (inner pinna), which had been previously characterized using Ion Torrent PGM™. We assessed skin microbiota from inner pinna samples: 1A, 7A, 18A, 20A and 29A from (36) (or Chapter 3.3).
Sample collection and DNA extraction
For the complex microbial community, skin microbiota samples were collected using Sterile Catch-All™ Sample Collection Swabs (Epicentre Biotechnologies) soaked in sterile SCF-1 solution (50 mM Tris buffer (pH = 8), 1 mM EDTA, and 0.5% Tween-20).
Bacterial DNA was extracted from the swabs using the PowerSoil™ DNA isolation kit (MO BIO) (for further details on sample collection and DNA extraction see (37)).
PCR amplification and barcoding
To prepare the DNA and the library we followed the Oxford Nanopore protocol 1D PCR barcoding amplicons (SQK-LSK108), however we used the Phusion Taq polymerase rather than the LongAmp Taq recommended in this protocol. Specifically, we amplified ~1,500bp fragments of the full 16S rRNA gene.
Bacterial DNA was amplified using a nested PCR with a first round to add the 16S rRNA gene primer sets and a second round to add the barcodes. In this study we used two sets of 16S universal primers. On one hand, primer set 27F-1391R (also named S-D-Bact-0008-c-S-20 and S-D-Bact-1391-a-A-17 (38)) amplified V1-V8 hypervariable regions of 16S rRNA gene. On the other hand, primer set 27F-1492R (also named S-D-Bact-0008-c-S-20 and S-D-Bact-1492-a-A-22 (38)) amplified V1-V9 hypervariable regions of 16S rRNA gene. These two sets of universal primers are the most commonly used when assessing full-length 16S rRNA gene, because they have shown a really low non-coverage rate, even at phylum level (39). The primers used in this study are listed in Table 1 and contain some ambiguous bases previously described to make the primers more universal (25).
Table 1. Primer sequences and hypervariable regions (HVR) targeted for full-length 16S rRNA gene amplification and sequencing.
Complete name Short name HVR Sequence (5' --> 3') Melting
T
S-D-Bact-0008-c-S-20 27F V1 AGRGTTYGATYMTGGCTCAG 54.4 ºC S-D-Bact-1391-a-A-17 1391R V8 GACGGGCGGTGWGTRCA 59.5 ºC S-D-Bact-1492-a-A-22 1492R V9 TACCTTGTTAYGACTT 41.6 ºC
We will distinguish among primer sets used referring to the hypervariable regions they are amplifying, so: 27F-338R will be V1-V2; 27F-1391R will be V1-V8; and 27F-1492R will be V1-V9.
We ordered the 16S rRNA gene primers with the Oxford Nanopore Universal Tag added to their 5’ end. The universal tag was 5’-TTTCTGTTGGTGCTGATATTGC-3’ for forward primers and 5’-ACTTGCCTGTCGCTCTATCTTC-3’ for reverse primers. These universal tags will allow the second barcoding PCR using the PCR Barcoding kit (EXP-PBC001).
In the first round, PCR mixture (25 μl) contained initial DNA sample (1 μl of DNA for the mock community PCR and 5 μl of DNA for the skin microbiota PCR), 5 μl of 5X Phusion Buffer HF, 0.2 mM of dNTPs, 0.02 U/μl of Phusion High Fidelity Taq Polymerase (Thermo Scientific). Primer concentrations were adapted to each primer set: for 27F-1391R, 0.4 μM of each primer and for 27F-1492R, 0.4μM of 27F and 0.8μM of 1492R. The PCR thermal profile consisted of an initial denaturation step for 30s at 98°C, followed by
25 cycles for 15s at 98°C, 15s at primer-adjusted annealing temperature, 45s at 72°C for extension, and a final step for 7 min at 72°C. The annealing temperature was also adjusted to the primer set: 55ºC for 27F-1391R and 51ºC for 27F-1492R. To assess possible reagent contamination, each PCR reaction included a no template control sample, which did not amplify.
In the second round, PCR mixture (100 μl) contained 0.5 nM of the first-round PCR product, 20 μl of 5X Phusion Buffer HF, 0.2 mM of dNTPs, 0.02 U/μl of Phusion High Fidelity Taq Polymerase (Thermo Scientific), and 2 μl of each specific barcode (EXP-PBC001) as recommended in the Oxford Nanopore protocol 1D PCR barcoding amplicons (SQK-LSK108). The PCR thermal profile consisted of an initial denaturation step for 30s at 98°C, followed by 15 cycles for 15s at 98°C, 15s at 62ºC for annealing, 45s at 72°C for extension, and a final extension step for 7 min at 72°C.
Following each PCR round, a clean-up step using AMPure XP beads at 0.5X concentration was used to discard short fragments as recommended by the manufacturer. DNA quantity was assessed using Qubit fluorimeter.
A final equimolar pool containing 1ug of the barcoded DNA samples in 45 uL of DNAse and RNAse free water will be used to prepare the sequencing library.
Library preparation
The Ligation Sequencing Kit 1D (SQK-LSK108) was used to prepare the amplicon library to load into the MinIONTM following the instructions of the 1D PCR barcoding amplicon protocol of ONT. Input DNA samples were1 μg of the barcoded DNA pool in a volume of 45 μL and 5 μL of DNA CS (DNA from lambda phage, used as a sequencing positive control). The DNA was processed for end repair and dA-tailing using the NEBNext End Repair / dA-tailing Module (New England Biolabs). A purification step using Agencourt AMPure XP beads (Beckman Coulter) was performed and approximately the expected 700ng of total DNA were recovered as assessed by Qubit quantification.
For the adapter ligation step, a total of 0.2 pmol of the end-prepped DNA (approximately 200 ng of our 1,500 bp fragment) were added in a mix containing 50μL of Blunt/TA ligase master mix (New England Biolabs) and 20μL of adapter mix, and were incubated at room temperature for 10 min. We performed a purification step using Agencourt AMPure XP beads (Beckman Coulter) and Adapter Bead Binding buffer provided on SQK-LSK108 kit to finally obtain the DNA library also called pre-sequencing mix.
Preparing and loading the flow cell
We used SpotON Flow Cell Mk I (R9.4) (FLO-MIN106), which had been previously stored at 4ºC. We fitted the flow cell to the MinIONTM and performed the Quality Control. We continue with the priming of the flowcell with a mixture of Running Buffer with fuel mix (RBF from SQK-LSK108) and Nuclease-free water (500 μL + 500 μL).
We prepared the pre-sequencing mix (12 μL of DNA library) to be loaded by mixing it with Library Loading beads (25.5 μL) and Running Buffer with fuel mix (37.5 μL). Immediately after priming, the nanopore sequencing library was loaded in a dropwise fashion using the spot-on port.
We run a total of two flow cells including several barcoded samples. The first one contained the mock communities amplified using V1-V9 primer set (M1 and M2 are technical replicates) together with other skin microbiota samples not included in this study. The second flow cell included the skin microbiota sample from the inner pinna amplified using V1-V8 and V1-V9 primer sets (biological replicates) described in this study, among others.
Once the library was loaded we initiated a standard 48h sequencing protocol using the MinKNOW™ software.
For the mock communities, basecalling was performed using the Metrichor™ agent 1D barcoding for pre-existing basecalls and demultiplexing using Epi2me debarcoding workflow. Finally, fast5 files were converted to fastq files using poRe (40)(Watson et al., 2015) and adapters were trimmed using Porechop (41).
On the other hand, for the skin microbiota samples basecalling was performed using Albacore v0.8.4 software. Again, fast5 files were converted to fastq files using poRe (40). Afterwards sequences were demultiplexed and adapters trimmed using Porechop (41).
As a final step, we trimmed the universal tags of the sequences using a custom script and filtered out those sequences shorter than 1,100 bp for V1-V8 and 1,200 bp for V1-V9 amplifications respectively, using split_libraries.py from QIIME software (42).
Downstream analyses
For both the mock community and the skin microbiota samples, we performed the analysis using NanoOK (43) with LAST aligner (44) against two databases. We used a subset of Greengenes database (45,46) and the rrn database (33).
Greengenes database offers annotated, chimera-checked, and full-length 16S rRNA gene sequences and it is one of the most commonly used databases when performing microbiota analyses(45,46). We used the Greengenes database clustered at 99% of similarity to reduce redundancy and filtered out those sequences that did not reach species level or that did not have a minimum length of 1,400bp. This database contained 20,745 sequences belonging to 3,147 different species.
rrn database is a custom database created by Benitez-Paez and Sanz to analyze the rrn operons. It contains information of the ribosomal RNA operon (16S-ITS-23S genes) retrieved from bacterial genomes of GenBank at NCBI. This database contained 22,351 sequences belonging to 2,384 different species (33).
Results and discussion
Mock community analyses
We amplified full-length 16S rRNA sequences from the staggered community with primers V1-V9 by duplicate (M1 and M2). We processed a total of 11,284 sequences for M1 and 22,995 for M2. The taxonomic results obtained are shown in Table 2, both for Greengenes and rrn databases.
Table 2. Taxonomic assignment of the mock community at species level. Results obtained after MinIONTM sequencing of two replicates of the staggered mock community (M1 and M2) aligned against Greengenes and rrn databases. Relative abundances that correlated to operon counts are in bold. (*) Is the species annotated in the database?
Greengenes database rrn database
Taxonomy nº of operons
Tax at sps?*
% reads M1
% reads M2
Tax at sps?*
% reads M1
% reads M2
Escherichia coli 1,000,000 Yes 15.29 16.45 Yes 22.49 23.92 Rhodobacter sphaeroides 1,000,000 Yes 8.54 7.19 Yes 8.18 6.85 Staphylococcus epidermidis 1,000,000 Yes 18.81 19.04 Yes 15.99 16.71 Streptococcus mutans 1,000,000 No - - Yes 17.34 17.12 Bacillus cereus 100,000 Yes 2.26 2.1 Yes 0.93 0.96 Clostridium beijerinckii 100,000 No - - Yes 0.46 0.67 Pseudomonas aeruginosa 100,000 Yes 0.01 0.03 Yes 0.85 0.88 Staphylococcus aureus 100,000 Yes 1.59 1.81 Yes 6.81 6.49 Streptococcus agalactiae 100,000 Yes 1.27 1.42 Yes 2.36 2.36 Acinetobacter baumannii 10,000 No - - Yes 0.11 0.09 Helicobacter pylori 10,000 Yes 0.1 0.14 Yes 0.11 0.13 Lactobacillus gasseri 10,000 No - - Yes 0.17 0.12 Listeria monocytogenes 10,000 Yes 0.21 0.29 Yes 0.12 0.23 Neisseria meningitides 10,000 No - - Yes 0.14 0.24 Propionibacterium acnes 10,000 Yes 0.17 0.22 Yes 0.03 0.05 Actinomyces odontolyticus 1,000 Yes - - Yes 0 0 Bacteroides vulgatus 1,000 No - - Yes 0.02 0.01 Deinococcus radiodurans 1,000 No - - No - - Enterococcus faecalis 1,000 No - - Yes 0.11 0.02 Streptococcus pneumoniae 1,000 No - - Yes 3.63 3.38
Greengenes is one of the most commonly used databases in microbiota studies because it is curated and checked for chimeras (46). However, half of the bacterial species of the mock community were not annotated in the database up to the species level, so we expected seeing only genus level (marked as “No” Table 2).
rrn database (33) contains information of the ribosomal RNA operon (16S-ITS-23S genes) for 22,351 sequences belonging to 2,384 different species. This database lacks information for only one member of the mock community (Deinococcus radiodurans).
On one hand, Greengenes contains taxonomic annotation up to the species level for 10 out of 20 bacterial species included in the mock bacterial community. From these, we were able to detect all of them. Moreover, it’s worthy to note that despite Streptococcus mutans was not in Greengenes database at species level, we detected the closely related species Streptococcus sobrinus in high abundance (M1=14.4% and M2=11.3%); in fact both belong to the mutans group (47). Moreover, we saw Streptococcus infantis (M1=9.6 and M2= 7.9%) as another abundant species. On the other hand, rrn database contained species level information from 19 out of the 20 bacterial species of the mock community and we were able to detect all of them (Table 2).
The overall trend when looking at relative abundances is that operon counts correlated to relative abundances: when operon count changed in one magnitude order also relative abundances did. The exceptions were Rhodobacter sphaeroides, Pseudomonas aeruginosa and Actinomyces odontolyticus, which were detected at lower abundances than expected or not detected at all.
When looking at the results of rrn database, we could see that not only Rhodobacter sphaeroides and Pseudomonas aeruginosa were underrepresented but also Bacillus cereus and Clostridium beijerinckii. Finally, Streptococcus pneumoniae was overrepresented, probably suggesting that the sequencing errors together with the large amount of Streptoccous entries in the rrn database (44 Streptococcus species in rrn vs 9 Streptococcus species in Greengenes) produced an incorrect identification of this species. Thus, some of these biases could be explained by sequencing errors, primers used, or low resolution of 16S rRNA gene to distinguish among species from certain genera.
At the genus level, we were able to identify all the members of the mock community except Actinomyces odontolyticus even when it was represented on both databases (Table 3). Aproximately 10% of the reads for Greengenes and 2% for rrn database belonged to other genera theoretically not present in the mock community. Among those “other genera”, the most abundant belonged to Shigella, Enterobacter and Salmonella, which are closely related to Escherichia coli (48). If we not consider these taxa probably wrongly assigned because they are closely related to Escherichia coli, only ~2.5% and ~0.5 % of the reads of Greengenes and rrn database respectively belong to other genera rather than the expected, which could be due to sequencing errors or to cross-contamination from dog skin microbiota samples.
Table 3. Taxonomic assignment of the mock community at genus level. Results obtained at the genus level after MinIONTM sequencing of two replicates of the staggered mock community (M1 and M2) aligned against Greengenes and rrn databases.
Greengenes rrn
Genus Expected abund.
% of reads M1
% of reads M2
% of reads M1
% of reads M2
Streptococcus ++++ 33.99 33.44 35.44 34.82 Staphylococcus ++++ 24.00 24.30 24.32 24.73 Escherichia ++++ 15.41 16.52 22.91 24.23 Rhodobacter ++++ 8.58 7.21 8.35 6.99 Bacillus +++ 3.66 3.69 3.28 3.37 Clostridium +++ 1.31 1.46 1.25 1.3 Pseudomonas +++ 1.04 1.05 1.07 1.01 Listeria ++ 0.27 0.30 0.20 0.30 Neisseria ++ 0.17 0.27 0.17 0.30 Propionibacterium ++ 0.17 0.22 0.13 0.17 Lactobacillus ++ 0.27 0.17 0.28 0.18 Helicobacter ++ 0.10 0.14 0.12 0.13 Acinetobacter ++ 0.13 0.12 0.13 0.11 Enterococcus + 0.14 0.11 0.12 0.06 Bacteroides + 0.02 0.01 0.02 0.01 Deinococcus + 0.02 0 0.02 0.01 Actinomyces + 0 0 0 0 Other genera* - 10.72 10.99 2.19 2.28
We can conclude from mock community analyses that full-length 16S rRNA sequencing with Oxford Nanopore is able to detect taxonomy assignments and retrieve diversity information, provided that the target species are in the database. At the genus level, we were able to accurately retrieve the mock community composition. It’s also worthy to note the good technical replicates obtained for M1 and M2 samples.
Evaluation of primer sets V1-V8 and V1-V9 in microbial richness
Dog skin microbiota samples were sequenced as a pool with MinIONTM after amplification of full-length 16S rRNA gene with primers targeting regions V1-V8 or V1-V9 (see Table 1). These complex microbiota samples were basecalled with Albacore v0.8.4 (Oxford Nanopore Technologies) and fast5 files were converted to fastq files. After demultiplexing, adapters and universal tags were trimmed and sequences analyzed using NanoOK (43) with LAST aligner (44). We finally obtained a total of 79,083 sequences for V1-V9 and 74,243 for V1-V8.
The same samples had been previously sequenced individually with Ion Torrent PGM™ with primers targeting V1-V2 hypervariable regions. In that case sequences were analyzed with QIIME 1.9.1 (42) with operational taxonomic units (OTUs) picking a representative
sequence of a group of sequences with a 97% similarity and taxonomy was assigned with RDP classifier (49) against the whole Greengenes database (45,46)(that contains many entries that do not reach low taxonomic levels). Using RDP classifier, if the taxonomy assignment does not reach a specific threshold, the sequences included in the OTU are set as “Other”. We finally obtained a total of 249,572 sequences for V1-V2 region.
We performed the evaluation of the primer sets using exclusively the Greengenes database, because V1-V2 short-reads were analyzed using this same database. We used a subset of the Greengenes database that contained only those taxa that reached species level for the long-reads obtained for V1-V8 and V1-V9 regions with MinIONTM. As a consequence of the stricter criteria for taxonomy assignment of V1-V2 short-reads, lower taxonomic levels were poorly annotated. Thus, we compared diversity estimates of higher taxa (from kingdom to order).
Both V1-V8 and V1-V9 primer sets for long-reads were able to retrieve more bacterial taxa than V1-V2 short reads. The bacterial richness was higher at different taxonomic levels when assessed with long-reads rather than with short-reads, as seen in Table 4, and this trend increased as we were lowering taxonomic level.
Table 4. Bacterial richness estimates for skin microbiota samples. Bacterial richness retrieved with different primer pairs targeting short (V1-V2) or full-length (V1-V8 and V1-V9) 16S rRNA gene.
V1-V2
(~350 bp) V1-V8
(~1300 bp) V1-V9
(~1400 bp) Kingdom 1 2 2 Phylum 16 22 22 Class 33 46 47 Order 53 88 91
At the highest taxonomic level, we were able to detect not only Bacteria but also Archaea kingdom, despite using universal primers specific for Bacteria (25). However, they were present at really low proportions (< 0.01% of total reads), which agrees with results previously reported on human skin microbiota samples (50).
Delving into phylum level, we detected that the most common and better characterized phyla were retrieved by all the primers. These taxa represented >98% of the total skin microbiota composition. Long-read primers were able to detect 8 phyla previously unseen using V1-V2 short-reads (Table 5). It has already been reported the low coverage of this primer set for some specific phyla (21,51). Some phyla were only detected with a specific primer set, such as Lentisphaerae with V1-V8 or Fibrobacteres with V1-V9. On the other hand, GN02, TM7 and Thermi phyla belong to candidate divisions and none of their members have been cultivated (52)(Camanocha et al., 2014), so databases do not have taxonomy information up to species level. Thus, since we used the Greengenes subset database with species-level sequences, no representative of those phyla were included for taxonomy assignment of V1-V8 and V1-V9 long-reads and that is probably the reason why they are only detected with V1-V2 primers.
Table 5. Microbial richness on skin samples. Table containing all the observed phyla per primer subset. Long-reads taxonomy was obtained from a species-level subset of Greengenes database (see materials and methods) that did not contain information of GN02, SR1 and TM7. *phyla that belong to Archaea kingdom. ** phyla with no representative at the spp level in the database used for assigning taxonomy to long-reads (V1-V8 and V1-V9).
Long-reads Short-reads
Phylum V1-V9 V1-V8 V1-V2 [Thermi] + + + Acidobacteria + + + Actinobacteria + + + Bacteroidetes + + + Chlorobi + + + Chloroflexi + + + Cyanobacteria + + + Deferribacteres + + + Firmicutes + + + Fusobacteria + + + Proteobacteria + + + Spirochaetes + + + Tenericutes + + + Aquificae + + Chlamydiae + + Crenarchaeota* + + Elusimicrobia + + Euryarchaeota* + + Planctomycetes + + Synergistetes + + Verrucomicrobia + + Fibrobacteres + Lentisphaerae + GN02** + SR1** + TM7** + Total 22/26 22/26 16/26
So, we were able to detect previously unseen bacteria phyla on dog skin using MinIONTM long-amplicons for full-length 16S rRNA sequences, which provided better richness estimates. These previously unseen bacteria phyla presented low relative abundances on canine skin microbiota, but the use of long-amplicons in more uncharacterized environments will provide better diversity estimates. In conclusion, as previously reported, the ability to classify sequences to the genus or species level is a function of read length, but also of sample type and the reference database (23).
Skin microbiota analyses
We could see from the mock community results that species-level resolution was a little bit tricky because sometimes the target species was not present in the database. However, here we aimed to compare the taxonomies at species level obtained from two independent databases containing different taxonomic annotations. We considered a species-level assignment valid when two unrelated databases gave identical annotation.
When comparing the most abundant species (>1% of total relative abundances), we could see 7 bacterial species identified by two independent databases: Bergeyella zoohelcum, Capnocytophaga canimorsus, Pasteurella multocida, Neisseria shayeganii, Lactobacillus reuteri, Bibersteinia trehalosi, and Neisseria weaver (Figure 1).
This study included 5 skin microbiota samples together in a pool, which had been previously sequenced individually using V1-V2 short-reads and Ion Torrent PGM™. Figure 2A shows the microbiota profile of the most abundant taxa per individual sample (taxa representing > 0.5%). We can see that all of them have been identified up to family level and some of them also to genus level (Figure 2A). Despite working with an equimolar pool of these samples for nanopore sequencing, we could see in Figure 2B that this experiment is not representing equally all the individual samples and is missing some taxa.
When comparing the results of nanopore sequencing with the short reads, we could detect that Lactobacillus species were exclusive to one single sample of the pool (18A), so probably not the most abundant taxa on the inner pinna skin microbiota. Fusobacteriaceae (specifically Fusobacterium genus) is one of the most abundant families on the inner pinna skin microbiota, and both databases detected that pattern at family level. However, Greengenes database was not able to identify any Fusobacterium, since the database lacked entries of the genus at the species level. Porphyromonadaceae was also another abundant member of skin microbiota, but species level taxonomy differed in each database giving different results. Long-reads allowed us reaching lower taxonomic levels confirmed by two independent databases for: 1) [Weeksellaceae] family, with Bergeyella zoohelcum; 2) Neisseriaceae, with Neisseria shayeganii and Neisseria weaveri; 3) Pasteurellaceae, with Pasteurella multocida and 4) Flavobacteriaceae, with Capnocitophaga canimorsus.
Figure 1. Skin microbiota composition at species level. Comparison of the species detected against the (a) Greengenes and (b) rrn databases in the dog skin microbiota samples amplified with V1-V8 and V1-V9 16S rRNA primers and sequenced with MinIONTM. (*) coincident taxonomic assignments using independent databases.
Figure 2. Skin microbiota taxonomic profile using long and short reads. Bar plots of (a) the most abundant taxa on individual skin samples sequenced with V1-V2 primers in short-reads. “g__” means there is no information at genus level; and b) main families on dog skin microbiota when sequencing the full-length 16S rRNA gene (V1-V9 and V1-V8) on a pool of samples and representation of the same families when sequencing V1-V2 region on the individual samples that constituted the pool.
Conclusions Nanopore sequencing of the full-length 16S rRNA gene allowed us inferring microbiota composition from both simple and complex microbial communities. Moreover, long-reads were able to retrieve increased richness estimates, which show us previously unseen phyla on dog skin, despite being at really low abundances.
Taxonomy assignment down to species level was not always feasible because of both the high error rate of 1D reads and the absence of some bacterial taxa on the databases. With the nanopore reads, we assigned taxonomy through alignment strategies executing all-vs-all comparisons that need many computational resources, so we needed a small database to obtain results. When working with a 16S database subset, we should be sure to include the most relevant taxa even if they do not have representative members at species level. Using only species-level taxonomy, we lost insights of candidate division phyla such as SR1, GN02 or TM7. Oxford nanopore offers other bioinformatics tools, such as the cloud-based EPI2ME platform. However, we run out of memory on our hard disk when trying to perform these analyses.
Finally, we should keep in mind that delving into species level with only 16S is sometimes difficult because there are bacterial species that share almost all their 16S rRNA sequences (60). Recently, Benitez-Paez and colleagues proposed to sequence the whole rrn operon constituted by 16S-ITS-23S to obtain better species-level resolution (33).
In conclusion, nanopore sequencing of the full-length 16S rRNA gene allowed identifying bacterial species even from a complex community, obtaining a microbiota profile and improving richness estimates. So, nanopore sequencing has the potential to be used to assess microbial communities. Future studies should be relying on the new 1D2 kit that presents higher accuracy and, since sequencing length is not an impediment, other experimental strategies could be assessed.
Bibliography 1. Marchesi JR, Ravel J. The vocabulary of microbiome research: a proposal. Microbiome
(2015) 3:31. doi:10.1186/s40168-015-0094-5
2. Costello EK, Lauber CL, Hamady M, Fierer N, Gordon JI, Knight R. Bacterial community variation in human body habitats across space and time. Science (2009) 326:1694–1697. doi:10.1126/science.1177486
3. The Human Microbiome Project Consortium. Structure, function and diversity of the healthy human microbiome. Nature (2012) 486:207–214. doi:10.1038/nature11234
4. Qin J, Li R, Raes J, Arumugam M, Burgdorf KS, Manichanh C, Nielsen T, Pons N, Levenez F, Yamada T, et al. A human gut microbial gene catalogue established by
metagenomic sequencing. Nature (2010) 464:59–65. doi:10.1038/nature08821
5. Belkaid Y, Segre JA. Dialogue between skin microbiota and immunity. Science (2014) 346:954–9. doi:10.1126/science.1260144
6. Thaiss CA, Zmora N, Levy M, Elinav E. The microbiome and innate immunity. Nature (2016) 535:65–74. doi:10.1038/nature18847
7. Belkaid Y, Hand TW. Role of the microbiota in immunity and inflammation. Cell (2014) 157:121–141. doi:10.1016/j.cell.2014.03.011
8. Fitz-Gibbon S, Tomida S, Chiu B-H, Nguyen L, Du C, Liu M, Elashoff D, Erfe MC, Loncaric A, Kim J, et al. Propionibacterium acnes strain populations in the human skin microbiome associated with acne. J Invest Dermatol (2013) 133:2152–60. doi:10.1038/jid.2013.21
9. Dreno B, Martin R, Moyal D, Henley JB, Khammari A, Seité S. Skin microbiome and acne vulgaris: staphylococcus, a new actor in acne. Exp Dermatol (2017) doi:10.1111/exd.13296
10. Gao Z, Tseng C, Strober BE, Pei Z, Blaser MJ. Substantial alterations of the cutaneous bacterial biota in psoriatic lesions. PLoS One (2008) 3:e2719. doi:10.1371/journal.pone.0002719
11. Fahlén A, Engstrand L, Baker BS, Powles A, Fry L. Comparison of bacterial microbiota in skin biopsies from normal and psoriatic skin. Arch Dermatol Res (2012) 304:15–22. doi:10.1007/s00403-011-1189-x
12. Alekseyenko A V, Perez-Perez GI, De Souza A, Strober B, Gao Z, Bihan M, Li K, Methé BA, Blaser MJ. Community differentiation of the cutaneous microbiota in psoriasis. Microbiome (2013) 1:31. doi:10.1186/2049-2618-1-31
13. Kong HH, Oh J, Deming C, Conlan S, Grice EA, Beatson MA, Nomicos E, Polley EC, Komarow HD, Murray PR, et al. Temporal shifts in the skin microbiome associated with disease flares and treatment in children with atopic dermatitis. Genome Res (2012). 22:850-9. doi:10.1101/gr.131029.111
14. Oh J, Freeman AF, Park M, Sokolic R, Candotti F, Holland SM, Segre JA, Kong HH. The altered landscape of the human skin microbiome in patients with primary immunodeficiencies. Genome Res (2013) 23:2103–2114. doi:10.1101/gr.159467.113
15. Seite S, Flores GE, Henley JB, Martin R, Zelenkova H, Aguilar L, Fierer N. Microbiome of affected and unaffected skin of patients with atopic dermatitis before and after emollient treatment. J Drugs Dermatol (2014) 13:1365–72.
16. Chng KR, Tay ASL, Li C, Ng AHQ, Wang J, Suri BK, Matta SA, McGovern N, Janela B, Wong XFCC, et al. Whole metagenome profiling reveals skin microbiome-dependent susceptibility to atopic dermatitis flare. Nat Microbiol (2016) 1:16106. doi:10.1038/nmicrobiol.2016.106
17. Kennedy EA, Connolly J, Hourihane JO, Fallon PG, McLean WHI, Murray D, Jo J-H, Segre JA, Kong HH, Irvine AD. Skin microbiome before development of atopic dermatitis: Early colonization with commensal staphylococci at 2 months is associated with a lower risk of atopic dermatitis at 1 year. J Allergy Clin Immunol (2017) 139:166–172. doi:10.1016/j.jaci.2016.07.029
18. Rodrigues Hoffmann A, Patterson AP, Diesel A, Lawhon SD, Ly HJ, Elkins Stephenson C, Mansell J, Steiner JM, Dowd SE, Olivry T, et al. The skin microbiome in healthy and allergic dogs. PLoS One (2014) 9:e83197. doi:10.1371/journal.pone.0083197
19. Bradley CW, Morris DO, Rankin SC, Cain CL, Misic AM, Houser T, Mauldin EA, Grice EA. Longitudinal evaluation of the skin microbiome and association with microenvironment and treatment in canine atopic dermatitis. J Invest Dermatol (2016) 136:1182-90. doi:10.1016/j.jid.2016.01.023
20. Pierezan F, Olivry T, Paps JS, Lawhon SD, Wu J, Steiner JM, Suchodolski JS, Hoffmann AR. The skin microbiome in allergen-induced canine atopic dermatitis. Vet Dermatol (2016) 27:332-e82. doi:10.1111/vde.12366
21. Kuczynski J, Lauber CL, Walters WA., Parfrey LW, Clemente JC, Gevers D, Knight R. Experimental and analytical tools for studying the human microbiome. Nat Rev Genet (2011) 13:47–58. doi:10.1038/nrg3129
22. Clarridge JE. Impact of 16S rRNA Gene Sequence Analysis for Identification of Bacteria on Clinical Microbiology and Infectious Diseases. Clin Microbiol Rev (2004) 17:840–862. doi:10.1128/CMR.17.4.840-862.2004
23. Schloss PD, Jenior ML, Koumpouras CC, Westcott SL, Highlander SK. Sequencing 16S rRNA gene fragments using the PacBio SMRT DNA sequencing system. PeerJ (2016) 4:e1869. doi:10.7717/peerj.1869
24. Yarza P, Yilmaz P, Pruesse E, Glöckner FO, Ludwig W, Schleifer K-H, Whitman WB, Euzéby J, Amann R, Rosselló-Móra R. Uniting the classification of cultured and uncultured bacteria and archaea using 16S rRNA gene sequences. Nat Rev Microbiol (2014) 12:635–645. doi:10.1038/nrmicro3330
25. Klindworth A, Pruesse E, Schweer T, Peplies J, Quast C, Horn M, Glockner FO. Evaluation of general 16S ribosomal RNA gene PCR primers for classical and next-generation sequencing-based diversity studies. Nucleic Acids Res (2013) 41:e1. doi:10.1093/nar/gks808
26. Fichot EB, Norman RS. Microbial phylogenetic profiling with the Pacific Biosciences sequencing platform. Microbiome (2013) 1:10. doi:10.1186/2049-2618-1-10
27. Mosher JJ, Bernberg EL, Shevchenko O, Kan J, Kaplan LA. Efficacy of a 3rd generation high-throughput sequencing platform for analyses of 16S rRNA. J
Microbiol Methods (2013) 95:175-81. doi: 10.1016/j.mimet.2013.08.009
28. Mosher JJ, Bowman B, Bernberg EL, Shevchenko O, Kan J, Korlach J, Kaplan LA. Improved performance of the PacBio SMRT technology for 16S rDNA sequencing. J Microbiol Methods (2014) 104:59–60. doi:10.1016/j.mimet.2014.06.012
29. Wagner J, Coupland P, Browne HP, Lawley TD, Francis SC, Parkhill J. Evaluation of PacBio sequencing for full-length bacterial 16S rRNA gene classification. BMC Microbiol (2016) 16:274. doi:10.1186/s12866-016-0891-4
30. Singer E, Bushnell B, Coleman-Derr D, Bowman B, Bowers RM, Levy A, Gies EA, Cheng J-F, Copeland A, Klenk H-P, et al. High-resolution phylogenetic microbial community profiling. ISME J (2016) 10:2020–2032. doi:10.1038/ismej.2015.249
31. Ma X, Stachler E, Bibby K. Evaluation of Oxford Nanopore MinION Sequencing for 16S rRNA Microbiome Characterization. bioRxiv (2017). doi:10.1101/099960
32. Benítez-Páez A, Portune KJ, Sanz Y. Species-level resolution of 16S rRNA gene amplicons sequenced through the MinIONTM portable nanopore sequencer. Gigascience (2016) 5:4. doi:10.1186/s13742-016-0111-z
33. Benitez-Paez A, Sanz Y. Multi-locus and ultra-long amplicon sequencing approach to study microbial diversity at species level using the MinION TM portable nanopore sequencer. bioRxiv (2017) doi:10.1101/117143
34. Mitsuhashi S, Kryukov K, Nakagawa S, Takeuchi JS, Shiraishi Y, Asano K, Imanishi T. A portable system for metagenomic analyses using nanopore-based sequencer and laptop computers can realize rapid on-site determination of bacterial compositions. bioRxiv (2017) https://doi.org/10.1101/101865
35. Shin J, Lee S, Go M-J, Lee SY, Kim SC, Lee C-H, Cho B-K. Analysis of the mouse gut microbiome using full-length 16S rRNA amplicon sequencing. Sci Rep (2016) 6:29681. doi:10.1038/srep29681
36. Cuscó A, Belanger JM, Gershony L, Islas-Trejo A, Levy K, Medrano JF, Sánchez A, Oberbauer AM, Francino O. Individual signatures and environmental factors shape skin microbiota on healthy dogs. Microbiome (submitted).
37. Cuscó A, Sánchez A, Altet L, Ferrer L, Francino O. Individual Signatures Define Canine Skin Microbiota Composition and Variability. Front Vet Sci (2017) 4:6. doi:10.3389/fvets.2017.00006
38. Alm E, Oerther D, Larsen N, Stahl D. The oligonucleotide probe database. Appl Environ Microbiol. (1996) 62:3557–3559
39. Mao D-P, Zhou Q, Chen C-Y, Quan Z-X. Coverage evaluation of universal bacterial primers using the metagenomic datasets. BMC Microbiol (2012) 12:66. doi:10.1186/1471-2180-12-66
40. Watson M, Thomson M, Risse J, Talbot R, Santoyo-Lopez J, Gharbi K, Blaxter M. PoRe: An R package for the visualization and analysis of nanopore sequencing data. Bioinformatics (2015) 31:114–115. doi:10.1093/bioinformatics/btu590
41. Wick R. Porechop. Available at: https://github.com/rrwick/Porechop
42. Caporaso JG, Kuczynski J, Stombaugh J, Bittinger K, Bushman FD, Costello EK, Fierer N, Peña AG, Goodrich JK, Gordon JI, et al. QIIME allows analysis of high-throughput community sequencing data. Nat Methods (2010) 7:335–336. doi:10.1038/nmeth.f.303
43. Leggett RM, Heavens D, Caccamo M, Clark MD, Davey RP. NanoOK: multi-reference alignment analysis of nanopore sequencing data, quality and error profiles. Bioinformatics (2015) 32:btv540. doi:10.1093/bioinformatics/btv540
44. Kielbasa SM, Wan R, Sato K, Horton P, Frith MC. Adaptive seeds tame genomic sequence comparison. Genome Res (2011) 21:487–493. doi:10.1101/gr.113985.110
45. DeSantis TZ, Hugenholtz P, Larsen N, Rojas M, Brodie EL, Keller K, Huber T, Dalevi D, Hu P, Andersen GL. Greengenes, a chimera-checked 16S rRNA gene database and workbench compatible with ARB. Appl Environ Microbiol (2006) 72:5069–72. doi:10.1128/AEM.03006-05
46. McDonald D, Price MN, Goodrich J, Nawrocki EP, DeSantis TZ, Probst A, Andersen GL, Knight R, Hugenholtz P. An improved Greengenes taxonomy with explicit ranks for ecological and evolutionary analyses of bacteria and archaea. ISME J (2012) 6:610–8. doi:10.1038/ismej.2011.139
47. Lal D, Verma M, Lal R. Exploring internal features of 16S rRNA gene for identification of clinically relevant species of the genus Streptococcus. Ann Clin Microbiol Antimicrob (2011) 10:28. doi:10.1186/1476-0711-10-28
48. Fukushima M, Kakinuma K, Kawaguchi R. Phylogenetic analysis of Salmonella, Shigella, and Escherichia coli strains on the basis of the gyrB gene sequence. J Clin Microbiol (2002) 40:2779–85. doi:10.1128/JCM.40.8.2779-2785.2002
49. Wang Q, Garrity GM, Tiedje JM, Cole JR. Naive Bayesian classifier for rapid assignment of rRNA sequences into the new bacterial taxonomy. Appl Environ Microbiol (2007) 73:5261–7. doi:10.1128/AEM.00062-07
50. Oh J, Byrd AL, Deming C, Conlan S, Barnabas B, Blakesley R, Bouffard G, Brooks S, Coleman H, Dekhtyar M, et al. Biogeography and individuality shape function in the human skin metagenome. Nature (2014) 514:59–64. doi:10.1038/nature13786
51. Walters WA, Caporaso JG, Lauber CL, Berg-Lyons D, Fierer N, Knight R. PrimerProspector: de novo design and taxonomic analysis of barcoded polymerase chain reaction primers. Bioinformatics (2011) 27:1159–61. doi:10.1093/bioinformatics/btr087
52. Camanocha A, Dewhirst FE. Host-associated bacterial taxa from Chlorobi, Chloroflexi, GN02, Synergistetes, SR1, TM7, and WPS-2 Phyla/candidate divisions. J Oral Microbiol (2014) 6:25468 doi:10.3402/jom.v6.25468
53. Wallis C, Marshall M, Colyer A, O’Flynn C, Deusch O, Harris S. A longitudinal assessment of changes in bacterial community composition associated with the development of periodontal disease in dogs. Vet Microbiol (2015) 181:271–282. doi:10.1016/j.vetmic.2015.09.003
54. Lin WR, Chen YS, Liu YC. Cellulitis and Bacteremia Caused by Bergeyella zoohelcum. J Formos Med Assoc (2007) 106:573–576. doi:10.1016/S0929-6646(07)60008-4
55. Oehler RL, Velez AP, Mizrachi M, Lamarche J, Gompf S. Bite-related and septic syndromes caused by cats and dogs. Lancet Infect Dis (2009) 9:439–447. doi:10.1016/S1473-3099(09)70110-0
56. Sturgeon A, Stull JW, Costa MC, Weese JS. Metagenomic analysis of the canine oral cavity as revealed by high-throughput pyrosequencing of the 16S rRNA gene. Vet Microbiol (2013) 162:891–898. doi:10.1016/j.vetmic.2012.11.018
57. Andersen BM, Steigerwalt AG, O’Connor SP, Hollis DG, Weyant RS, Weaver RE, Brenner DJ. Neisseria weaveri sp. nov., formerly CDC group M-5, a gram-negative bacterium associated with dog bite wounds. J Clin Microbiol (1993) 31:2456–66.
58. Kim SY, Adachi Y. Biological and genetic classification of canine intestinal lactic acid bacteria and bifidobacteria. Microbiol Immunol (2007) 51:919–28.
59. Silva BC, Jung LRC, Sandes SHC, Alvim LB, Bomfim MRQ, Nicoli JR, Neumann E, Nunes AC. In vitro assessment of functional properties of lactic acid bacteria isolated from faecal microbiota of healthy dogs for potential use as probiotics. Benef Microbes (2013) 4:267–75. doi:10.3920/BM2012.0048
60. Janda JM, Abbott SL. 16S rRNA gene sequencing for bacterial identification in the diagnostic laboratory: pluses, perils, and pitfalls. J Clin Microbiol (2007) 45:2761–4. doi:10.1128/JCM.01228-07
131
4. Discussion
This thesis intends to shed light onto the skin microbiota on healthy dogs as well as
providing a tool to assess genetic variability in innate immunity.
In the research article “Non-synonymous genetic variation in exonic regions of canine Toll-like
receptors” we characterized by massive sequencing the genetic polymorphisms of Toll-like
receptors on dogs from seven different breeds and on wolves from two populations.
Finally we developed a TaqMan OpenArray® panel with 64 non-synonymous SNPs with a
likely functional effect on the TLR proteins and we validated it using a subset of the
previously sequenced animals.
In the research article “Individual signatures define canine skin microbiota composition and variability”
we characterized the variability of the skin microbiota by analyzing eight different skin sites
in a cohort of nine healthy dogs from three different breeds: German Shepherd, French
Bulldog and West Highland White Terrier. We assessed factors such as breed, skin site and
individual as potential microbiota drivers. We found that the individual was the main factor
driving skin microbiota structure and composition, followed by the skin site. We clarified
that this individual effect should be understood as the dog, its lifestyle, and its
environment. Finally, we detected no effect of the breed, however only three animals per
breed were used so this effect should be better assessed.
To resolve if that individual effect shaping skin microbiota on healthy dogs came from the
dog itself or from the environment we performed a third research article “Individual
signatures and environmental factors shape skin microbiota on healthy dogs”. Here we characterized
the variability of the skin microbiota in healthy dogs cohabiting together in a shared
environment, by analyzing eight different skin sites in a cohort of thirty-five Golden-
Labrador Retriever crossbred dogs. Even for this cohort with a shared environment, we
found that microbiota composition was driven by the individual, confirming the results of
132
the second research article. When considering abundances, the microbiota structure was
driven both by the individual and by the skin site. Considering that the cohort was
environmentally uniform, we were able to detect the effect of host-specific factors: the
season when the animal was born or the time spent in the kennel was a factor shaping skin
microbiota in all skin sites.
Finally in the chapter “Using MinION™ to characterize dog skin microbiota through full-length 16S
rRNA gene sequencing approach” we assessed the potential of Nanopore sequencing for
microbiota analyses using the full-length 16S rRNA gene. We first sequenced a simple
microbial mock community containing 20 bacterial species (HM-783D) and then a
complex microbial community represented by a pool of different skin microbiota samples
from the inner pinna. Despite using 1D chemistry, which presents low accuracy, we were
able to obtain better richness estimates and some information at species level. However, we
should consider using the new 1D2 chemistry from now on, and exploiting the potential of
this technology performing other approaches.
In each of these four chapters the obtained results are thoroughly discussed. Thus, the
purpose of the present section is to review and unify the four independent studies, as well
as to discuss the pitfalls and to propose future directions.
133
4.1. Dual assessment of innate immunity and
skin microbiota
In the research article “Non-synonymous genetic variation in exonic regions of canine Toll-like
receptors” (Chapter 3.1), we characterized by massive sequencing the genetic variability in the
coding regions of the 10 Toll-like Receptor (TLRs) genes in 355 dogs from 7 breeds and in
100 wolves from 2 populations. We functionally annotated the different variants and
predicted the likely effect of non-synonymous SNPs (nsSNPs) on the TLR proteins that
act as the first sensors of microbes. We found that the frequencies of nsSNPs differed
among breeds and that some of them were breed- or species-specific.
Finally, we developed a TaqMan OpenArray genotyping plate with 60 non-synonymous
SNPs and 4 frameshift mutations. This tool allows obtaining an innate immunity profile on
single dogs by genotyping their TLR polymorphisms. It can be used in many areas, such as
characterizing genetic variability on different dog breeds (Chapter 3.1) and even performing
evolution studies when including other canid species; screening case-control cohorts to
identify variants associated to a certain disease; and finally, characterizing innate immune
profile of individual dogs that are also analyzed in a microbiota study.
When we screened the genetic variability of TLRs on several dogs from different breeds,
we found that Labrador and German Shepherd dogs clustered separately (Figure 28a).
Thus, some breeds present different innate immune profiles that could be influencing their
predisposition and outcome to certain diseases, as well as their microbiota structure and
composition.
Three of the nsSNPs from TLR5 gene identified by massive sequencing had been
previously associated with Inflammatory Bowel diseases in German Shepherd (Kathrani et
al., 2010) and in other breeds (Kathrani et al., 2011). In our cohort of healthy dogs, we
found that the risk allele for IBD in German Shepherd was absent in this breed although it
presented different frequencies in other breeds (Figure 28b). TLR5 is the innate immune
receptor for flagellin that is the principal protein component of bacterial flagella (Leifer et
al., 2014). German Shepherd dogs carrying this risk allele for canine IBD showed hyper-
responsiveness to flagellin and Kathrani and colleagues suggested that this altered innate
immunity-microbiota cross-talk could be responsible for the inappropriate inflammation
observed in this disease (Kathrani et al., 2012). More recently, Vazquez-Baeza and
collaborators confirmed that dogs with IBD presented an altered fecal microbiota that in
fact could be used as a diagnosis tool to distinguish them from the healthy individuals
(Vázquez-Baeza et al., 2016). Similarly on human IBD, Knights and collaborators reported
a significant association between risk alleles of the innate immune receptor NOD2 and an
increased relative abundance of Enterobacteriaceae in gastrointestinal microbiota (Knights et
al., 2014).
134
Figure 28. Genetic variation on canine Toll-like Receptors. In a) Principal Component Analysis of dogs individually genotyped using TaqMan Open Array plate. In b) allele frequencies in our cohort of non-synonymous SNPs obtained through massive sequencing and associated to IBD (From Chapter 3.1, Cuscó et al., 2014).
Regarding the aim of this dissertation, this tool will allow us to perform a dual assessment
of the dog in health and disease: innate immune profile and microbiota composition.
In healthy skin, microbiota and immunity interact to maintain the homeostasis in front of
disruptions. In fact, mice studies have demonstrated impaired and weakened skin immune
responses in Germ free mice when compared to Pathogen-Specific Free mice (Naik et al.,
2012b). In healthy human skin, some of these cross-talks between specific commensals and
innate immunity have already been characterized (Barnard and Li, 2017; Holmes et al.,
2015; Nakamizo et al., 2015). Staphylococcus epidermidis is one of the most studied
commensals and interacts with host cell’s TLRs to modulate innate immunity against
pathogens (Table 3, Figure 24). Other commensals such as Propionibacterium acnes,
Pseudomonas aeruginosa or Staphylococcus aureus contribute to skin immunity by producing
antimicrobial substances (Holmes et al., 2015).
In dermatological diseases, the equilibrium and cross-talks between skin immunity and
microbiota are disrupted. As we have extensively reviewed in the introduction, most of the
cutaneous diseases are associated with an altered microbiota or dysbiosis (Table 4), and/or
an altered innate immunity, either through genetic polymorphisms or altered expression of
TLRs (Table 2). When comparing the two tables, some common cutaneous diseases appear
135
to present both an altered microbiota and an altered skin immunity profile. As an example,
some links between disease and TLR-microbiota have been merged in Table 5.
Table 5. TLR-microbiota associations in main cutaneous diseases.
Disease TLR-microbiota link
Atopic dermatitis - During AD flare S. aureus increases, but also S. epidermidis which might be a
compensatory mechanism (see Table 3 for specific mechanisms)
- Topical probiotic of Vitreoscilla filiformis reduced AD score, increasing IL-10
production via TLR2 (Volz et al., 2014)
Psoriasis - Polymorphisms on innate immune system receptor (e.g. TLRs) facilitate the
initiation of an inflammatory response to commensal microbiota (Fry et al.,
2015)
Acne vulgaris - P. acnes strains are individual-specific, some virulent and some not (Fitz-
Gibbon et al., 2013)
- Health-associated phylotypes produce thiopeptides, which are antimicrobial
compounds that inhibit the growth of gram-positive bacteria (Christensen and
Brüggemann, 2014)
Some correlations between host genetic variation and certain microbiota composition have
already been described in health and disease. Knights and colleagues, as we have previously
explained, found a link between innate immune receptor NOD2 and increased
Enterobacteriaceae on gastrointestinal microbiota. Moreover, they suggested that genetically
altered host functional pathways can shape microbiome structure (Knights et al., 2014).
Blekham and colleagues approach was retrieving the host sequences of the shotgun
metagenomics dataset of the Human Microbiome Project. First they correlated host
functional pathways (groups of aggregated SNPs and genes) with overall microbiome
composition and found significant enrichment in several genes involved in (i) complex
diseases that have already been linked to the microbiome (such as obesity and
Inflammatory Bowel disease); and (ii) other immunity-related pathways, including the Role
of Pattern Recognition Receptors in Recognition of Bacteria and Viruses. They also
correlated host variation to specific microbiome members and found key host genes related
to immunity involved. For example, genetic variants on HLA-DRA and TLR1 were
correlated to Selenomonas presence on throat microbiota and Lautropia in the tongue
dorsum, respectively (Blekhman et al., 2015).
To sum up, TLRs serve the dual function of sensing pathogens and symbionts and this
sensing leads to very different outcomes for both microbes (clearance vs. symbiosis) and
host (inflammation vs. immune homeostasis) (Chu and Mazmanian, 2013). Genetic variants
on these genes can affect these responses. Thus, future uses of the canine TLR chip with
the non-synonymous SNPs could be in both health and disease: in health, to elucidate if
some specific TLR variants are associated to specific microbiota compositions; and in
disease, to elucidate if some specific TLR variants are associated to some specific diseases
and maybe in turn with an altered microbiota composition.
136
Finally, we should not forget the third piece of this puzzle: the environment. An excellent
example of interaction between host genome, microbiome and environment is observed in
humans by the interaction among LCT gene, Bifidobacterium and diet (dairy consumption).
Lactose is the main sugar of dairy products and can be metabolized in the gastrointestinal
tract either by Bifidobacterium of the microbiota or the lactase enzyme of the host, which is
codified by LCT gene. Lactose intolerance in adults is associated with the SNP rs4988235
with genotype GG in LCT gene. Only in GG carriers, Bifidobacterium abundance is
correlated with dairy consumption suggesting that microbiota is replacing and/or
complementing the host genetics function (Figure 29) (Bonder et al., 2016). Thus,
microbiota can adapt to different host genetics backgrounds to complement functions
using environmental resources.
Figure 29. Example of an interaction between host genome, microbiota and environment: LCT gene, Bifidobacterium and dairy consumption. In individuals that present the GG genotype in LCT gene, the dairy consumption is positively correlated with the presence of Bifidobacterium in the GI tract (excerpted from Bonder et al., 2016).
Taken together, these findings motivate the need for larger association studies to
characterize host genetic variation linked to the microbiome in the context of various
health conditions, environmental effects, and genetic backgrounds.
137
4.2. Skin site signatures on canine skin
microbiota
Human skin is mainly divided in three microhabitats depending on its physicochemical
properties that harbor specific microbiota: sebaceous sites, with Propionibacterium spp;
moist sites with Staphylococcus and Corynebacterium spp; and dry sites, with gram-negative
microorganisms (Costello et al., 2009; Grice and Segre, 2011; Grice et al., 2009b).
Dog skin is more uniform and almost totally covered by a dense fur, and despite the
anatomical and physicochemical differences among certain skin sites (reviewed at Chapter
1.2) these are not as large as in humans. Moreover, environment has probably a
homogenizing effect hiding some skin-site bacterial signatures. Despite the more
homogeneous microhabitat, skin diseases present different prevalence depending on skin
sites (Miller et al., 2013).
Even with this uniformity, the first cross-sectional study of skin microbiota on healthy
dogs identified that hairy skin presented higher diversity values when compared to mucosal
surfaces or muco-cutaneous junctions, as well as skin-site specific taxonomic profiles
(Rodrigues Hoffmann et al., 2014). We have expanded this knowledge, performing two
more cross-sectional studies on healthy dogs and obtaining similar results for diversity, with
low diversity values in perianal and nasal skin microbiota samples (Chapter 3.2 and 3.3,
Cuscó et al., 2017a, 2017b).
Identifying skin-site bacterial signatures is biologically meaningful, and reflects the
anatomical and physicochemical differences present on dog skin (Figure 30). We found
influences of gastrointestinal and oral microbiota on the perianal region and the chin,
reflecting their physical proximity (Chapter 3.2 and 3.3, Cuscó et al., 2017a, 2017b). These
other body site microbiotas present higher bacterial concentrations than the skin (Belkaid
and Segre, 2014a; Sender et al., 2016). The inner pinna presented the most uniform skin
microbiota composition among individuals despite being the most diverse skin site when
compared to the other skin sites, probably because it was the most isolated part of the skin
of the dogs included in our studies (Chapter 3.2 and 3.3, Cuscó et al., 2017a, 2017b).
To obtain first insights on the functional potential of a microbial community, predictive
tools that work with 16S rRNA gene data, such as PICRUSt (Langille et al., 2013), PanFP
(Jun et al., 2015) or Tax4fun (Aßhauer et al., 2015) have proven to be useful. At the
predicted functional level, we were able to detect an increase on lipid metabolism on back
skin, where sebaceous glands are larger and more abundant (Figure 30c). This last finding
suggested that microbiota was adapting to the specific microhabitat or resource.
138
Figure 30. Microbiota of healthy dogs depending on the skin site. In a) microbiota profile at phylum level of the 9 dogs cohort. In b) boxplots of the diversity values depending of each skin site, marked with an asterisk those significantly different skin sites. In c) differentially distributed predicted functions on different skin sites (Figures from Cuscó et al., 2017a).
Further studies assessing the metagenomes, using shotgun whole genome sequencing, will
allow detecting rather than predicting the functional potential of the microbial community.
However, metagenomics studies on the skin are especially challenging due to the low
biomass obtained from this tissue (Kong et al., 2017).
To conclude, we have detected different skin site microbial signatures that elucidate the
need of cross-sectional studies and the avoidance of general results or even therapies
focused on “the skin” as a unique microhabitat. Specifically, we found few examples on
how skin site properties shape its specific microbiota, even one at the functional level. The
next step will be finding out how these specific microbial signatures contribute to skin
functions to promote health or disease.
139
4.3. Individual signatures on skin microbiota
Inter-individual diversity is high in human skin microbiota (Grice and Segre, 2011; Human
Microbiome Project Consortium, 2012) and can even be used to identify individuals by
matching them to objects that the person came in contact with (Fierer et al., 2010; Lax et
al., 2015; Meadow et al., 2014). Generally low-abundant microorganisms are the ones with
the highest identification power, tend to be spread over all the skin sites, and their
abundances remain relatively constant through time (Oh et al., 2016).
Despite prior studies suggested individuality was affecting canine skin microbiota
(Rodrigues Hoffmann et al., 2014), they did not assess this factor directly. Here, we
specifically found that the individual was the main factor shaping dog skin microbiota
composition: samples resembled more each other within the same dog, even when
including different skin sites (Figure 31) (Chapter 3.2 and 3.3, Cuscó et al., 2017a, 2017b).
In Chapter 3.2, we could not distinguish if that effect was purely the individual or maybe it
was due to their unique associated environment (the 9 dogs came from different
households). However, in Chapter 3.3 we worked with a large cohort of same-breed dogs
cohabiting together in the same environment and our results confirmed that the individual
was the main factor shaping skin microbiota.
Figure 31. Skin microbiota profile of healthy dogs at phylum level. A, B, C, D, E, F, G and H represent inner pinna, chin, nasal skin, dorsal back, axilla, abdomen, interdigital region and perianal region. (Figure modified from Cuscó et al., 2017a).
140
This individuality of the skin microbiota should be taken into account on disease
treatments. Skin microbiota is altered during several cutaneous diseases on humans and on
atopic dermatitis on dogs (reviewed at Section 1.3.4). Considering that fact together with
host genetics, generic therapies are probably not the best option. These individual effects
could explain the differences on efficacy of the different treatments for skin diseases, such
as those seen for canine atopic dermatitis (Olivry and Bizikova, 2013).
Microbiota manipulation can be potentially used to treat diseases as reviewed previously on
Section 1.3.5. Using a skin microbiota survey, Bradley and colleagues found Staphylococcus
and Corynebacterium increased in canine atopic dermatitis (Figure 32a). Despite these general
characteristics, they treated the affected dogs with antimicrobials and see a highly
personalized trend on both the degree of improvement (Figure 32b, Visit 2) and the
outcome after antibiotics removal (Figure 32b, Visit 3) (Bradley et al., 2016b).
Figure 32. Dog skin microbiota on atopic dermatitis. a) Microbiota profile at genus level on healthy and atopic dermatitis groups. b) Atopic dermatitis score through time: visit 1, is the initial point; visit 2, during antimicrobial treatment; visit 3, after removing antimicrobials (Figures from Bradley et al., 2016).
Besides classical antimicrobial administration, other therapies such as administration of pre-
and probiotics could be used to treat or prevent certain diseases. In the light of our results,
where the individual is the main factor shaping dog skin microbiota, assessing individual
microbiota profile for each dog would help to understand the different outcomes to a
common treatment between individuals. The study from Bradley and colleagues
demonstrates the necessity to assess the individual microbial signatures and to use this
information to develop personalized therapies (Bradley et al., 2016b).
These individual signatures are often seen at low taxonomic levels, such as species level
(Oh et al., 2014). To achieve that, we assessed 3rd generation sequencing technologies using
MinION™ and found that species-level assignment was possible in some cases when
sequencing the full-length 16S rRNA gene (Chapter 3.4).
These individual-specific characteristics of skin microbiota have been seen not only at
compositional level, but also in the dynamics of microbial populations using network
studies. Bashan and colleagues found that among the skin sites tested (left and right
antecubital fossae and retro-auricular creases), only retro-auricular creases skin microbiota
presented universal dynamics in humans (Bashan et al., 2016).
141
In conclusion, individuality of dog skin microbiota is a fact that veterinary clinicians and
researchers should account for, especially when developing new treatments to modify or
alter skin microbiota. It is worthy to note that individuality is not synonymous of host
genetics neither host environment, but it is probably a combination of many factors and it
remains to be clarified to which extent one affect to the other. Despite this individuality, it
will also be interesting to assess if any universal interactions exist in specific skin sites, as
those seen in retro-auricular creases in human (Bashan et al., 2016). If microbial dynamics
of dog skin are not universal, generic microbiota manipulations may result ineffective or
even detrimental.
142
4.4. Environmental bacteria on skin: transient
or resident microbiota?
A recent study on caterpillars has demonstrated that they lack a resident gut microbiota,
and they only have bacteria that come from the environment and the ingested food, with
more than 80% of the 16S rRNA reads belonging to chloroplasts. Treating the animals
with antibiotics did not affect their fitness or survival, demonstrating that the transient
microbiota was not contributing on their overall health (Hammer et al., 2017).
In a relatively isolated environment such as the gut, it is easier to control whether the
microbiota comes from the environment or not. However that is much more difficult in
the skin, which is totally exposed to the outer environment. For example, tribal populations
that live mostly outdoors and have close contact with nature presented an incredible rich
microbiota with many environmental-derived bacteria, when compared to westernized
civilizations (Clemente et al., 2015; Hospodsky et al., 2014).
In Chapter 3.2 and 3.3 we assessed skin microbiota of healthy dogs and detected a clear
influence of the environment, both by the chloroplast amplification (ranging from 2 to
77% of the total sequences per sample in Chapter 3.2) and by the environmental-associated
bacteria such as Xenococcaceae, Sphingomonadaceae or Pseudomonadaceae (Cuscó et al.,
2017a)(Cuscó et al., 2017b). Previous studies also detected environmental-derived bacteria
on canine skin microbiota (Rodrigues Hoffmann et al., 2014; Song et al., 2013a; Torres et
al., 2017).
Song and colleagues suggested that dogs not only harbor a resident microbiota but also
shed a transient one (Song et al., 2013a). Even some built environment microbiota studies
considered dogs as vectors of the outer environment to the household, increasing indoors-
bacterial diversity (Kettleson et al., 2015; Miletto and Lindow, 2015).
The large proportion of chloroplasts sequences seen on some dog skin microbiota samples
probably showed their close and recent contact with vegetation and could be indicating
that some of the bacteria that we detected on these samples come from plants (Figure 33).
Comparing low-chloroplast with high-chloroplast skin microbiota samples after
chloroplasts removal will give powerful insights in identifying environment-associated
microbiota. Moreover in future studies it will be interesting to sample not only the dog skin
but also the environment (the soil where it walks, the plants which it interacts with, the
water where it plays, etc.).
143
Figure 33. Dogs interacting with the environment.
In conclusion, we detected that a variable proportion of the skin microbiota on dogs came
from the outer environment. Temporal studies through seasons can give some clues of
environmental-derived taxa (Torres et al., 2017), but if they interact or not with the host
microbial dynamics and the resident microbiota remains to be elucidated and it will
probably be harder to determine. Using some of the approaches proposed here could help
to give light to this question.
144
4.5. Conducting microbiota studies: 16S rRNA
gene and beyond
Depending on the research question to answer, alternative experimental scenarios can be
used besides targeting specific 16S rRNA gene hypervariable regions and sequencing them
using 2nd generation sequencers. In Figure 34, we provide an overview of the main
experimental approaches to perform a microbiota study and their main benefits and
pitfalls.
Figure 34. Experimental approaches to perform a microbiota study. (Source: own preparation)
145
Amplicon-based approach targeting specific hypervariable regions of 16S rRNA gene using
second generation sequencing is the preferred choice when conducting microbiota studies.
This approach, despite working with short fragments, allows high sequencing depth, low
error rates and affordable prices (depending on the technology chosen). Moreover, a broad
range of bioinformatics’ tools are freely available to analyze the data. Thus, the main
approach used on this dissertation will be valid for future analyses.
The most commonly used primers for microbiota studies are F27–R338 that amplify V1-
V2 hypervariable regions and F515–R806 that amplify V4 (Kuczynski et al., 2011). In this
thesis we choose the primer set F27-R338 (V1-V2), because despite F515–R806 (V4)
primer set is more universal (Walters et al., 2011), it lacks sensitivity for Propionibacterium
acnes, which is a common commensal of human skin microbiota (Figure 35)(Grice et al.,
2009).
Figure 35. Taxonomic coverage at phylum level for bacteria of two universal 16S rRNA gene primer sets. In a) the 27F-338R primer set (used in this study) and in b) the 515F/806R primer set. The y-axes represent percent coverage and the value on top of each bar is the total number of reference sequences in each taxon (adapted from Walters et al., 2011).
146
On the other hand, when working with other biological samples rather than skin,
researchers should choose the most appropriate primer set. In GI microbiota the most
suitable primer set is F515–R806 (V4), considering that F27–R338 (V1-V2) lacks sensitivity
for Bifidobacterium or the whole phylum Verrumicrobia, both expected to be in the GI tract
(Kuczynski et al., 2011). For example, the primer set used in this thesis is not able to
identify Akkermansia muciniphila (from Verrumicrobia phylum), one of the species with the
highest anti-inflammatory effects that is being studied as a potential probiotic. Moreover, it
is associated with weight-loss and lean phenotypes and also its relative abundance on GI
microbiota is lower in several metabolic diseases, such as obesity or inflammatory bowel
diseases, when compared to health status (Belzer and de Vos, 2012; Derrien et al., 2017;
Gómez-Gallego et al., 2016).
Therefore, in the light of our results (Chapter 3.2 and Chapter 3.3), where Propionibacterium
species represented really low abundances of the total skin microbiota on healthy dogs, it
would also be interesting to replicate the analyses with F515–R806 (V4) primer set,
especially in cases where an anti-inflammatory effect is studied.
When analyzing short-reads, most bioinformatics approaches rely on operational
taxonomic units (OTUs). Working with OTUs helps to reduce data volume and facilitates
the posterior analyses by clustering together all the sequences above a specific threshold
(usually 97%). As it has to process all the sequences, this step is usually slow and is
considered as the bottleneck on microbiota data analyses. This methodology can present
several disadvantages, such as 1) irreproducibility, OTU clustering algorithms lead to
divergent result from a same dataset (Schmidt et al., 2015); 2) instability, sequences forming
an OTUs vary depending on the number of sequences in the analysis, unless clustering
against a database (He et al., 2015); 3) inaccuracy, OTU diversity estimates can be inflated
due to sequencing errors or PCR artifacts (Patin et al., 2013). Recently, new algorithms
have been developed aimed at improving velocity and accuracy of OTU picking, such as
OptiClust (Westcott and Schloss, 2017), Deblur (Amir et al., 2017) or DADA2 (Callahan et
al., 2016) among many others.
To sum up, if using second generation sequencing and 16S rRNA gene, the hypervariable
region targeted and the universal primer sets used should be carefully chosen depending on
the sample of interest. Moreover, using the most recently released software and applying
strict quality filters will allow obtaining more accurate and reliable results.
As we have seen in chapters 3.2 and 3.3, taxonomy assignments of short fragments of 16S
rRNA obtained through second generation massive sequencing do not reach species level.
The main reasons are that: i) fragment length analyzed cannot discriminate among several
bacterial species and full 16S rRNA gene is required (Drancourt et al., 2000; Janda and
Abbott, 2007; Schloss et al., 2016a; Yarza et al., 2014); ii) analyses of microbiota data rely
on OTU-based clustering approaches, where all the sequences that are similar above a
specific threshold (usually 97%) are considered the same species altering diversity estimates
(He et al., 2015; Patin et al., 2013; Rossi-Tamisier et al., 2015); and iii) RDP taxonomy
147
classifier only computes reliable assignments up to genus level (Schloss et al., 2016a; Wang
et al., 2007).
Third generation sequencing generates longer reads that allow targeting the full 16S rRNA
gene and obtaining better richness estimates and taxonomic resolution (Yarza et al., 2014).
In fact, the ability to classify a sequence to the genus or species level is a function of read
length, sample type and the reference database (Schloss et al., 2016a). Thus, to overcome
some of the short-read associated problems, we tested 3rd generation MinION™ platform
to sequence the full 16S rRNA gene. On one hand, we were able to obtain information
down to species level, as well as we did see increased bacterial richness. On the other hand,
this was not possible in all the cases because of the databases and data analyses used were
not optimum, and the chemistry used (1D reads) still presented low accuracy values.
Moreover, the single 16S rRNA gene has high sequence similarity and not enough
taxonomic resolution for some species.
Benitez-Paez and Sanz suggested another approach to improve taxonomic resolution using
nanopore sequencing by targeting not only 16S rRNA gene, but the whole ribosomal RNA
operon (rrn region) constituted by 16S-ITS-23S (~4,500 bp). This approach allowed
obtaining a better taxonomic resolution at species level as well as seeing 2-fold more
diversity than when using only 16S rRNA gene (Benitez-Paez and Sanz, 2017).
Some of the pitfalls associated with the sequencing technology are being improved by
Oxford Nanopore Technologies. They have recently launched the 1D2 sequencing
chemistry (May 2017), which improves the accuracy of the reads by sequencing both
template and complement strands (Figure 36). To improve accuracy, Intramolecular-ligated
Nanopore Consensus Sequencing (INC-Seq) approach could be applied to microbiota
studies. INC-Seq uses rolling circle amplification of circularized templates to generate linear
products (with tandem copies of the template) that can be sequenced on the nanopore
platform (Li et al., 2016). Besides that, several bioinformatics tools and strategies have been
suggested to further correct the raw sequences, such as NanoCORR (Goodwin et al., 2015)
that corrects with an hybrid approach combining the long nanopore reads with short high-
quality reads or Nanocorrect and Nanopolish (Loman et al., 2015) that use self-correction
algorithms. Some genome assemblers, such as canu (Koren et al., 2017) also perform
previous steps of correcting and consensus sequences that can improve accuracy. These
tools are mostly developed for de novo genome assembly, and take advantage of the fact that
you are sequencing one genome and some regions overlap, which is not so useful for
amplicon-based microbiota studies. Further studies of microbiota should be using 1D2
sequencing kit together with suitable correction methods.
148
Figure 36. Evolution on the accuracy of Oxford Nanopore Technologies sequencing kits (Figure from ONT, Brown, 2017).
Sometimes 16S rRNA marker gene is not enough. It has several disadvantages such as: i)
“universal primers” used to amplify this gene were designed using sequences obtained from
cultured microorganisms, which could be biasing the actual diversity (Rajendhran and
Gunasekaran, 2011; Schloss and Handelsman, 2004); ii) databases are ecosystem-skewed
(Schloss et al., 2016b); 3) the copy number of rRNA operons per bacterial genome varies
from 1 to 15 (Klappenbach et al., 2001), which can lead to inflated bacterial diversity
estimates for some species; and 4) some species present a high similarity in this gene, as we
have corroborated in Chapter 3.4 for Streptococcus mutans or for Escherichia coli.
To overcome the first issue, Karst and colleagues developed a primer-free amplification of
SSU rRNA based on tagging single cDNA molecules obtained from size-selected rRNA
(aimed to enrich SSU rRNA). When comparing to SILVA database, they found that 30%
of all bacterial OTUs were novel and that the degree of novelty was highly ecosystem
specific, ranging from 36% in soil sample to 5% in human gut sample (Karst et al., 2016).
This approach elucidated that databases are environmental-skewed (Figure 37), and the
authors suggested that creating, maintaining and using an environment-specific 16S rRNA
gene database would be more feasible and appropriate to analyze specific
microhabitats(Karst et al., 2016), such as The Human Oral Microbiome Database (Chen et
al., 2010).
149
Figure 37. Coverage of the tree of life. The percent identity of SSU rRNA gene sequences in the samples compared to their closest relatives in the SILVA database. Red, green, and blue represent archaea, bacteria, and eukarya, respectively (Figure from Karst et al., 2016).
To overcome the third and fourth problems, which are more linked to the biological
properties of 16S rRNA gene, some researchers rely on another universal barcode for
bacteria: cpn60 protein coding gene. This gene is usually present as a single copy on
bacteria, which would improve relative abundance profiles. As exceptions to this general
rule Chlamydia, and some members of Rhizobia and Actinobacteria, present multiple copies of
the gene (Lund, 2009) (Figure 38a), whereas several species of Mycoplasma and Ureaplasma
parvum lack this gene (Hill et al., 2004). Links and colleagues compared the two bacterial
barcodes under the framework of the International Barcode of Life project and found that
the most informative regions of the 16S rRNA gene (V1-V3 region) are less taxonomically
informative than the most conserved segments of the cpn60 UT, for which average
sequence identity does not exceed 92% (Figure 38b). Thus, cpn60 UT has more taxonomic
resolution and is able to reach species and even subspecies level (Links et al., 2012).
Moreover, this marker gene also has a curated reference database: cpnDB (Hill et al., 2004).
Another approach would be sequencing RNA directly, which is now possible with
nanopore sequencing (Garalde et al., 2016) and even has been optimized for 16S rRNA
specifically (Smith et al., 2017). Some divergences at strain level can be consequence of
base modifications on the 16S rRNA. For example, 16S ribosomal RNA methylation has
been associated to resistance against aminoglycosides on several bacterial species (Doi and
Arakawa, 2007). Using direct RNA sequencing on MinION, Smith and colleagues were
able to detect the modified bases of Escherichia coli, which can identify a pathogenic strain
(Smith et al., 2017). Thus, this approach has potential to be used on microbiota studies to
assess taxonomy without PCR bias as well as to detect epigenetic divergences among
bacterial strains.
150
Figure 38. Comparison of universal bacterial barcodes: 16S rRNA and Cpn60. In a) pie charts representing the number of bacterial barcode gene copies per bacterial genome; in parenthesis number of genomes included per taxa. In b) sequence diversity across the 16S rRNA gene and cpn60 UT. Figure a) data from Větrovský and Baldrian, 2013 for 16S and Lund, 2009 for cpn60. Figure b) Excerpted from Links et al., 2012.
The broadest approach to study the microbiota is whole-genome sequencing that allows
getting the complete picture by sequencing all microbial members, including bacteria,
archaea, fungi, viruses or even eukaryotes. This approach aims to sequence all the genomes
in a specific environment and not only a marker gene, which means getting information of
the functional potential of the community as well as strain level identification (Kong et al.,
2017; Meisel et al., 2016). Few studies on human skin have already used it (Chng et al.,
2016; Oh et al., 2014, 2016). The main challenge of this technique is data analysis
complexity and host-associated DNA contamination. Moreover, skin samples present low
DNA biomass, which could add more contamination problems associated with laboratory
reagents or laboratory environment (Kong et al., 2017; Salter et al., 2014). Moreover, on
dog skin we would also expect to see contamination from chloroplasts DNA, as we have
already detected with 16S rRNA (Chapter 3.2 and 3.3, Cuscó et al., 2017a, 2017b).
Thus, plenty of options have already been proposed and assessed to characterize
microbiota structure and composition (Figure 34) and in this thesis we focused on two of
them. In the near future, new technologic and bioinformatics advances will allow creating
and developing new strategies and some of the new approaches stated here will become
routine. It is always important to keep in mind that depending on the research question to
answer, some scenarios are more suitable than others.
151
5. Conclusions
Regarding the innate immunity on healthy dogs, we can conclude that:
1. Genetic variation on TLR genes is greater than previously reported, we have
identified 105 novel coding variants by massive sequencing on different dog breeds
and wolf populations.
2. Frequencies of TLR genetic variants differ among dog breeds and wolf
populations, even some of them are breed-specific or species-specific.
3. The extracellular TLR5 is the most polymorphic among all the canine TLR genes.
4. We designed and validated a TaqMan OpenArray plate containing 60 non-
synonymous variants and 4 frameshift mutations spread in the 10 canine TLR
genes.
5. Individual genotyping of some dogs representing different breeds allowed us
detecting that Labrador and to lesser extent German Shepherd breeds presented
different innate immune profiles, when screening TLR genes.
Regarding the skin microbiota on healthy dogs, we can conclude that:
6. The individual is the main factor shaping skin microbiota: different skin sites within
the same dog resemble more than the same skin site between different dogs. That is
true even in a homogeneous environment.
7. The skin site is defining skin microbiota structure and composition. The inner
pinna (hairy skin) is the skin site with the highest diversity, whereas nasal skin and
the perianal region (muco-cutaneous regions) are the skin sites with the lowest.
8. The season of birth or the time spent in the kennel was affecting dog skin
microbiota in all the skin sites of dogs cohabiting together for at least 3 months.
Sphingomonas was the most abundant bacterium driving this difference.
152
9. Bacteria from the environment are highly present on dog skin microbiota, although
whether they are transient or permanent members of the microbiota remains to be
elucidated.
10. Other factors such as breed, sex and surgery seem to be affecting canine skin
microbiota, although larger studies should be performed to obtain stronger
evidence.
Regarding the assessment of single molecule sequencing experimental approach to obtain
microbiota data, we can conclude that:
11. Nanopore sequencing of full-length 16S rRNA gene using MinION™ allowed
obtaining better richness estimates and detecting previously unseen taxa on dog
skin microbiota.
12. Nanopore sequencing of full-length 16S rRNA gene using MinION™ allowed
obtaining taxonomy assignment at lower levels, even at species level in some cases.
153
6. Bibliography
Abida, O., Bouzid, D., Krichen-Makni, S., Kharrat, N., Masmoudi, A., Abdelmoula, M., Ben Ayed, M., Turki,
H., Sellami-Boudawara, T., and Masmoudi, H. (2013). Potential Role of TLR Ligand in Aethiopathogenesis of
Tunisian Endemic Pemphigus Foliaceus. Biochem. Physiol. Open Access 2, 1–5.
Agarwal, S.K., Wu, M., Livingston, C.K., Parks, D.H., Mayes, M.D., Arnett, F.C., and Tan, F.K. (2011). Toll-
like receptor 3 upregulation by type I interferon in healthy and scleroderma dermal fibroblasts. Arthritis Res.
Ther. 13, R3.
Al-Bagdadi, F. (2013). The Integument. In Miller’s Anatomy of the Dog, H.E. Evans, and A. de Lahunta, eds.
(Elsevier Inc.), 61–79.
Albertsen, M., Karst, S.M., Ziegler, A.S., Kirkegaard, R.H., Nielsen, P.H., and Stokholm-Bjerregaard, M.
(2015). Back to Basics – The Influence of DNA Extraction and Primer Choice on Phylogenetic Analysis of
Activated Sludge Communities. PLoS One 10, e0132783.
Alekseyenko, A. V, Perez-Perez, G.I., De Souza, A., Strober, B., Gao, Z., Bihan, M., Li, K., Methé, B.A., and
Blaser, M.J. (2013). Community differentiation of the cutaneous microbiota in psoriasis. Microbiome 1, 31.
Allenspach, K., House, A., Smith, K., McNeill, F.M., Hendricks, A., Elson-Riggins, J., Riddle, A., Steiner,
J.M., Werling, D., Garden, O.A., et al. (2010). Evaluation of mucosal bacteria and histopathology, clinical
disease activity and expression of Toll-like receptors in German shepherd dogs with chronic enteropathies.
Vet. Microbiol. 146, 326–335.
Amir, A., McDonald, D., Navas-Molina, J.A., Kopylova, E., Morton, J.T., Zech Xu, Z., Kightley, E.P.,
Thompson, L.R., Hyde, E.R., Gonzalez, A., et al. (2017). Deblur Rapidly Resolves Single-Nucleotide
Community Sequence Patterns. mSystems 2, e00191-16
Arumugam, M., Raes, J., Pelletier, E., Paslier, D., Yamada, T., Mende, D.R., Fernandes, G.R., Tap, J., Bruls,
T., Batto, J., et al. (2011). Enterotypes of the human gut microbiome. Nature 473, 174–180.
Ashelford, K.E., Chuzhanova, N.A., Fry, J.C., Jones, A.J., and Weightman, A.J. (2005). At least 1 in 20 16S
rRNA sequence records currently held in public repositories is estimated to contain substantial anomalies.
Appl. Environ. Microbiol. 71, 7724–7736.
Aßhauer, K.P., Wemheuer, B., Daniel, R., and Meinicke, P. (2015). Tax4Fun: predicting functional profiles
from metagenomic 16S rRNA data. Bioinformatics 31, 2882–2884.
Bakry, O.A., Samaka, R.M., Sebika, H., and Seleit, I. (2014). Toll-like receptor 2 and P. acnes: do they trigger
initial acne vulgaris lesions? Anal. Quant. Cytopathol. Histopathol. 36, 100–110.
Bangert, C., Brunner, P.M., and Stingl, G. (2011). Immune functions of the skin. Clin. Dermatol. 29, 360–
376.
Baquerizo Nole, K.L., Yim, E., and Keri, J.E. (2014). Probiotics and prebiotics in dermatology. J. Am. Acad.
Dermatol. 71, 814–821.
Barnard, E., and Li, H. (2017). Shaping of cutaneous function by encounters with commensals. J. Physiol.
595, 437–450.
154
Barreiro, L.B., Ben-Ali, M., Quach, H., Laval, G., Patin, E., Pickrell, J.K., Bouchier, C., Tichit, M., Neyrolles,
O., Gicquel, B., et al. (2009). Evolutionary dynamics of human Toll-like receptors and their different
contributions to host defense. PLoS Genet. 5, e1000562.
Bashan, A., Gibson, T.E., Friedman, J., Carey, V.J., Weiss, S.T., Hohmann, E.L., and Liu, Y.-Y. (2016).
Universality of human microbial dynamics. Nature 534, 259–262.
Bastos, M.C.F., Ceotto, H., Coelho, M.L. V, and Nascimento, J.S. (2009). Staphylococcal antimicrobial
peptides: relevant properties and potential biotechnological applications. Curr. Pharm. Biotechnol. 10, 38–61.
Begon, E., Michel, L., Flageul, B., Beaudoin, I., Jean-Louis, F., Bachelez, H., Dubertret, L., and Musette, P.
(2007). Expression, subcellular localization and cytokinic modulation of Toll-like receptors (TLRs) in normal
human keratinocytes: TLR2 up-regulation in psoriatic skin. Eur. J. Dermatol. 17, 497–506.
Belkaid, Y., and Segre, J.A. (2014). Dialogue between skin microbiota and immunity. Science 346, 954–959.
Belzer, C., and de Vos, W.M. (2012). Microbes inside—from diversity to function: the case of Akkermansia.
ISME J. 6, 1449–1458.
Benitez-Paez, A., and Sanz, Y. (2017). Multi-locus and ultra-long amplicon sequencing approach to study
microbial diversity at species level using the MinION TM portable nanopore sequencer. bioRxiv.
Beutler, B. (2004). Inferences, questions and possibilities in Toll-like receptor signalling. Nature 430, 257–263.
Beutler, B. (2009). Microbe sensing, positive feedback loops, and the pathogenesis of inflammatory diseases.
Immunol. Rev. 227, 248–263.
Blekhman, R., Goodrich, J.K., Huang, K., Sun, Q., Bukowski, R., Bell, J.T., Spector, T.D., Keinan, A., Ley,
R.E., Gevers, D., et al. (2015). Host genetic variation impacts microbiome composition across human body
sites. Genome Biol. 16, 1–12.
Bochud, P.Y., Sinsimer, D., Aderem, A., Siddiqui, M.R., Saunderson, P., Britton, S., Abraham, I., Tadesse
Argaw, A., Janer, M., Hawn, T.R., et al. (2009). Polymorphisms in Toll-like receptor 4 (TLR4) are associated
with protection against leprosy. Eur. J. Clin. Microbiol. Infect. Dis. 28, 1055–1065.
Bonder, M.J., Kurilshikov, A., Tigchelaar, E.F., Mujagic, Z., Imhann, F., Vila, A.V., Deelen, P., Vatanen, T.,
Schirmer, M., Smeekens, S.P., et al. (2016). The effect of host genetics on the gut microbiome. Nat. Genet.
48, 1407–1412.
Bradley, C.W., Morris, D.O., Rankin, S.C., Cain, C.L., Misic, A.M., Houser, T., Mauldin, E.A., and Grice,
E.A. (2016). Longitudinal evaluation of the skin microbiome and association with microenvironment and
treatment in canine atopic dermatitis. J. Invest. Dermatol. 136, 1182-90.
Brown, C. (2017). Nanopore technology update, London Calling conference 2017.
Burgener, I.A., König, A., Allenspach, K., Sauter, S.N., Boisclair, J., Doherr, M.G., and Jungi, T.W. (2008).
Upregulation of Toll-Like Receptors in Chronic Enteropathies in Dogs. J. Vet. Intern. Med. 22, 553–560.
Cadieu, E., Neff, M.W., Quignon, P., Walsh, K., Chase, K., Parker, H.G., VonHoldt, B.M., Rhue, A., Boyko,
A., Byers, A., et al. (2009). Coat Variation in the Domestic Dog Is Governed by Variants in Three Genes.
Science. 326, 150–153.
Callahan, B.J., McMurdie, P.J., Rosen, M.J., Han, A.W., Johnson, A.J.A., and Holmes, S.P. (2016). DADA2:
High-resolution sample inference from Illumina amplicon data. Nat. Methods 13, 581–583.
Callewaert, C., Kerckhof, F., Granitsiotis, M.S., Gele, M. Van, Wiele, T. Van De, and Boon, N. (2013).
Characterization of Staphylococcus and Corynebacterium Clusters in the Human Axillary Region. PLoS One.
8, e70538.
155
Capone, K.A., Dowd, S.E., Stamatas, G.N., and Nikolovski, J. (2011). Diversity of the human skin
microbiome early in life. J. Invest. Dermatol. 131, 2026–2032.
Caporaso, J.G., Bittinger, K., Bushman, F.D., Desantis, T.Z., Andersen, G.L., and Knight, R. (2010).
PyNAST: A flexible tool for aligning sequences to a template alignment. Bioinformatics 26, 266–267.
Caporaso, J.G., Kuczynski, J., Stombaugh, J., Bittinger, K., Bushman, F.D., Costello, E.K., Fierer, N., Peña,
A.G., Goodrich, J.K., Gordon, J.I., et al. (2010). QIIME allows analysis of high-throughput community
sequencing data. Nat. Methods 7, 335–336.
Castelino, M., Eyre, S., Moat, J., Fox, G., Martin, P., Ho, P., Upton, M., and Barton, A. (2017). Optimisation
of methods for bacterial skin microbiome investigation: primer selection and comparison of the 454 versus
MiSeq platform. BMC Microbiol. 17, 23.
Celhar, T., Magalhães, R., and Fairhurst, A.M. (2012). TLR7 and TLR9 in SLE: when sensing self goes wrong.
Immunol. Res. 53, 58–77.
Chakravorty, S., Helb, D., Burday, M., and Connell, N. (2007). A detailed analysis of 16S ribosomal RNA
gene segments for the diagnosis of pathogenic bacteria. J Microbiol Methods 69, 330–339.
Chen, T., Yu, W.-H., Izard, J., Baranova, O. V., Lakshmanan, A., and Dewhirst, F.E. (2010). The Human
Oral Microbiome Database: a web accessible resource for investigating oral microbe taxonomic and genomic
information. Database 2010, baq013.
Chng, K.R., Tay, A.S.L., Li, C., Ng, A.H.Q., Wang, J., Suri, B.K., Matta, S.A., McGovern, N., Janela, B.,
Wong, X.F.C.C., et al. (2016). Whole metagenome profiling reveals skin microbiome-dependent susceptibility
to atopic dermatitis flare. Nat. Microbiol. 1, 16106.
Chotimanukul, S., and Sirivaidyapong, S. (2011). Differential expression of Toll-like receptor 4 (TLR4) in
healthy and infected canine endometrium. Theriogenology 76, 1152–1161.
Christensen, G.J.M., and Brüggemann, H. (2014). Bacterial skin commensals and their role as host guardians.
Benef. Microbes 5, 201–215.
Chu, H., and Mazmanian, S.K. (2013). Innate immune recognition of the microbiota promotes host-microbial
symbiosis. Nat Immunol 14, 668–675.
Clarke, J., Wu, H.-C., Jayasinghe, L., Patel, A., Reid, S., and Bayley, H. (2009). Continuous base identification
for single-molecule nanopore DNA sequencing. Nat Nano 4, 265–270.
Clarke, K.R. (1993). Non-parametric multivariate analyses of changes in community structure. Aust. J. Ecol.
18, 117–143.
Clarridge, J.E. (2004). Impact of 16S rRNA Gene Sequence Analysis for Identification of Bacteria on Clinical
Microbiology and Infectious Diseases. Clin. Microbiol. Rev. 17, 840–862.
Clarys, P., and Barel, A. (1995). Quantitative evaluation of skin surface lipids. Clin. Dermatol. 13, 307–321.
Clemente, J.C., Pehrsson, E.C., Blaser, M.J., Sandhu, K., Gao, Z., Wang, B., Magris, M., Hidalgo, G.,
Contreras, M., Noya-Alarcon, O., et al. (2015). The microbiome of uncontacted Amerindians. Sci. Adv. 1,
e1500183.
Clooney, A.G., Fouhy, F., Sleator, R.D., and Driscoll, A.O. (2016). Comparing Apples and Oranges?: Next
Generation Sequencing and Its Impact on Microbiome Analysis. PLoS One. 11, e0148028.
Cogen, A.L., Yamasaki, K., Muto, J., Sanchez, K.M., Crotty Alexander, L., Tanios, J., Lai, Y., Kim, J.E., Nizet,
V., and Gallo, R.L. (2010). Staphylococcus epidermidis Antimicrobial δ-Toxin (Phenol-Soluble Modulin-γ)
Cooperates with Host Antimicrobial Peptides to Kill Group A Streptococcus. PLoS One 5, e8557.
156
Cole, J.R., Wang, Q., Fish, J.A., Chai, B., McGarrell, D.M., Sun, Y., Brown, C.T., Porras-Alfaro, A., Kuske,
C.R., and Tiedje, J.M. (2014). Ribosomal Database Project: data and tools for high throughput rRNA analysis.
Nucleic Acids Res. 42, D633–D642.
Costello, E.K., Lauber, C.L., Hamady, M., Fierer, N., Gordon, J.I., and Knight, R. (2009). Bacterial
community variation in human body habitats across space and time. Science 326, 1694–1697.
Cotton, D., van Hasselt, P., and Bergers, A. (1975). Nature of the sweat glands in the hairy skin of the beagle.
Dermatologica 150, 75–85.
Cuscó, A., Belanger, J.M., Gershony, L., Islas-Trejo, A., Levy, K., Medrano, J.F., and Sánchez, Armand
Oberbauer, Anita M. Francino, O. (2017b). Individual signatures and environmental factors shape skin
microbiota on healthy dogs. Microbiome (submitted).
Cuscó, A., Sánchez, A., Altet, L., Ferrer, L., and Francino, O. (2017a). Individual Signatures Define Canine
Skin Microbiota Composition and Variability. Front. Vet. Sci. 4, 6.
Derrien, M., Belzer, C., and de Vos, W.M. (2017). Akkermansia muciniphila and its role in regulating host
functions. Microb. Pathog. 106, 171–181.
DeSantis, T.Z., Hugenholtz, P., Larsen, N., Rojas, M., Brodie, E.L., Keller, K., Huber, T., Dalevi, D., Hu, P.,
and Andersen, G.L. (2006). Greengenes, a chimera-checked 16S rRNA gene database and workbench
compatible with ARB. Appl. Environ. Microbiol. 72, 5069–5072.
Dimarzio, L., Cinque, B., Cupelli, F., De Simone, C., Cifone, M.G., and Giuliani, M. (2008). Increase of Skin-
Ceramide Levels in Aged Subjects following a Short-Term Topical Application of Bacterial Sphingomyelinase
from Streptococcus thermophilus. Int. J. Immunopathol. Pharmacol. 21, 137–143.
Doi, Y., and Arakawa, Y. (2007). 16S Ribosomal RNA Methylation: Emerging Resistance Mechanism against
Aminoglycosides. Clin. Infect. Dis. 45, 88–94.
Dominianni, C., Wu, J., Hayes, R.B., and Ahn, J. (2014). Comparison of methods for fecal microbiome
biospecimen collection. BMC Microbiol. 14, 103.
Drancourt, M., Bollet, C., Carlioz, A., Martelin, R., Gayral, J.P., and Raoult, D. (2000). 16S ribosomal DNA
sequence analysis of a large collection of environmental and clinical unidentifiable bacterial isolates. J. Clin.
Microbiol. 38, 3623–3630.
Dreno, B., Martin, R., Moyal, D., Henley, J.B., Khammari, A., and Seité, S. (2017). Skin microbiome and acne
vulgaris: Staphylococcus, a new actor in acne. Exp. Dermatol. 1-6.
Dressman, D., Yan, H., Traverso, G., Kinzler, K.W., and Vogelstein, B. (2003). Transforming single DNA
molecules into fluorescent magnetic particles for detection and enumeration of genetic variations. Proc. Natl.
Acad. Sci. U.S.A. 100, 8817–8822.
Eckburg, P.B., Lepp, P.W., and Relman, D.A. (2003). Archaea and their potential role in human disease.
Infect. Immun. 71, 591–596.
Edgar, R.C., Haas, B.J., Clemente, J.C., Quince, C., and Knight, R. (2011). UCHIME improves sensitivity and
speed of chimera detection. Bioinformatics 27, 2194–2200.
Egert, M., and Simmering, R. (2016). The Microbiota of the Human Skin. In Advances in Experimental
Medicine and Biology, Adv Exp Med Biol. 902, 61–81.
Eid, J., Fehr, A., Gray, J., Luong, K., Lyle, J., Otto, G., Peluso, P., Rank, D., Baybayan, P., Bettman, B., et al.
(2009). Real-Time DNA Sequencing from Single Polymerase Molecules. Science 323, 133–138.
157
Fahlén, A., Engstrand, L., Baker, B.S., Powles, A., and Fry, L. (2012). Comparison of bacterial microbiota in
skin biopsies from normal and psoriatic skin. Arch. Dermatol. Res. 304, 15–22.
Farina, G.A., York, M.R., Di Marzio, M., Collins, C.A., Meller, S., Homey, B., Rifkin, I.R., Marshak-
Rothstein, A., Radstake, T.R, and Lafyatis, R. (2010). Poly(I:C) Drives Type I IFN- and TGFβ-Mediated
Inflammation and Dermal Fibrosis Simulating Altered Gene Expression in Systemic Sclerosis. J. Invest.
Dermatol. 130, 2583–2593.
Faust, K., and Raes, J. (2016). CoNet app: inference of biological association networks using Cytoscape.
F1000Research 5, 1519.
Faye, T., Holo, H., Langsrud, T., Nes, I.F., and Brede, D.A. (2011). The unconventional antimicrobial
peptides of the classical propionibacteria. Appl. Microbiol. Biotechnol. 89, 549–554.
Ferrer, M., Méndez-García, C., Rojo, D., Barbas, C., and Moya, A. (2017). Antibiotic use and microbiome
function. Biochem. Pharmacol. 134, 114–126.
Ferrer-Admetlla, A., Bosch, E., Sikora, M., Marquès-Bonet, T., Ramírez-Soriano, A., Muntasell, A., Navarro,
A., Lazarus, R., Calafell, F., Bertranpetit, J., et al. (2008). Balancing selection is the main force shaping the
evolution of innate immunity genes. J. Immunol. 181, 1315–1322.
Fierer, N., Hamady, M., Lauber, C.L., and Knight, R. (2008). The influence of sex, handedness, and washing
on the diversity of hand surface bacteria. Proc. Natl. Acad. Sci. U.S.A. 105, 17994–17999.
Fierer, N., Lauber, C.L., Zhou, N., McDonald, D., Costello, E.K., and Knight, R. (2010). Forensic
identification using skin bacterial communities. Proc. Natl. Acad. Sci. U.S.A. 107, 6477–6481.
Figueiredo, M.M., Amorim, I.F.G., Pinto, A.J.W., Barbosa, V.S., Pinheiro, L.D.J., Deoti, B., Faria, A.M.C.,
and Tafuri, W.L. (2013). Expression of Toll-like receptors 2 and 9 in cells of dog jejunum and colon naturally
infected with Leishmania infantum. BMC Immunol. 14, 22.
Findley, K., Oh, J., Yang, J., Conlan, S., Deming, C., Meyer, J.A., Schoenfeld, D., Nomicos, E., Park, M.,
Kong, H.H., et al. (2013). Topographic diversity of fungal and bacterial communities in human skin. Nature
498, 367–370.
Fischer, M., and Ehlers, M. (2008). Toll-like Receptors in Autoimmunity. Ann. N. Y. Acad. Sci. 1143, 21–34.
Fisher, R.A. (1936). The use of multiple measurements in taxonomic problems. Ann. Eugen. 7, 179–188.
Fitz-Gibbon, S., Tomida, S., Chiu, B.-H., Nguyen, L., Du, C., Liu, M., Elashoff, D., Erfe, M.C., Loncaric, A.,
Kim, J., et al. (2013). Propionibacterium acnes strain populations in the human skin microbiome associated
with acne. J. Invest. Dermatol. 133, 2152–2160.
Flores, G.E., Caporaso, J., Henley, J.B., Rideout, J., Domogala, D., Chase, J., Leff, J.W., Vázquez-Baeza, Y.,
Gonzalez, A., Knight, R., et al. (2014). Temporal variability is a personalized feature of the human
microbiome. Genome Biol. 15, 531.
Fouhy, F., Clooney, A.G., Stanton, C., Claesson, M.J., and Cotter, P.D. (2016). 16S rRNA gene sequencing of
mock microbial populations- impact of DNA extraction method, primer choice and sequencing platform.
BMC Microbiol. 16, 123.
Foulongne, V., Sauvage, V., Hebert, C., Dereure, O., Cheval, J., Gouilh, M.A., Pariente, K., Segondy, M.,
Burguière, A., Manuguerra, J.C., et al. (2012). Human skin microbiota: high diversity of DNA viruses
identified on the human skin by high throughput sequencing. PLoS One 7, e38499.
Fry, L., Baker, B.S., Powles, A. V., and Engstrand, L. (2015). Psoriasis is not an autoimmune disease? Exp.
Dermatol. 24, 241–244.
158
Fuentes, S., van Nood, E., Tims, S., Heikamp-de Jong, I., ter Braak, C.J., Keller, J.J., Zoetendal, E.G., and de
Vos, W.M. (2014). Reset of a critically disturbed microbial ecosystem: faecal transplant in recurrent
Clostridium difficile infection. ISME J. 8, 1621–1633.
Gallo, R.L., and Nakatsuji, T. (2011). Microbial symbiosis with the innate immune defense system of the skin.
J. Invest. Dermatol. 131, 1974–1980.
Ganju, P., Nagpal, S., Mohammed, M., Nishal Kumar, P., Pandey, R., Natarajan, V.T., Mande, S.S., and
Gokhale, R.S. (2016). Microbial community profiling shows dysbiosis in the lesional skin of Vitiligo subjects.
Sci. Rep. 6, 18761.
Gao, Z., Tseng, C., Strober, B.E., Pei, Z., and Blaser, M.J. (2008). Substantial alterations of the cutaneous
bacterial biota in psoriatic lesions. PLoS One 3, e2719.
Garalde, D.R., Snell, E.A., Jachimowicz, D., Heron, A.J., Bruce, M., Lloyd, J., Warland, A., Pantic, N.,
Admassu, T., Ciccone, J., et al. (2016). Highly parallel direct RNA sequencing on an array of nanopores.
bioRxiv.
Gay, N.J., and Gangloff, M. (2007). Structure and Function of Toll Receptors and Their Ligands. Annu. Rev.
Biochem. 76, 141–165.
Georgel, P., Macquin, C., and Bahram, S. (2009). The heterogeneous allelic repertoire of human toll-like
receptor (TLR) genes. PLoS One 4, e7803.
Gevers, D., Cohan, F.M., Lawrence, J.G., Spratt, B.G., Coenye, T., Feil, E.J., Stackebrandt, E., de Peer, Y.
Van, Vandamme, P., Thompson, F.L., et al. (2005). Opinion: Re-evaluating prokaryotic species. Nat. Rev.
Microbiol. 3, 733–739.
Glenn, T.C. (2011). Field guide to next-generation DNA sequencers. Mol. Ecol. Resour. 11, 759–769.
Gómez-Gallego, C., Pohl, S., Salminen, S., De Vos, W.M., and Kneifel, W. (2016). Akkermansia muciniphila:
a novel functional microbe with probiotic properties. Benef. Microbes 7, 571–584.
Gontcharova, V., Youn, E., Sun, Y., Wolcott, R.D., and Dowd, S.E. (2010). A comparison of bacterial
composition in diabetic ulcers and contralateral intact skin. Open Microbiol. J. 4, 8–19.
Goodwin, S., Gurtowski, J., Ethe-Sayers, S., Deshpande, P., Schatz, M.C., and McCombie, W.R. (2015).
Oxford Nanopore sequencing, hybrid error correction, and de novo assembly of a eukaryotic genome.
Genome Res. 25, 1750–1756.
Goodwin, S., McPherson, J.D., and McCombie, W.R. (2016). Coming of age: ten years of next-generation
sequencing technologies. Nat. Rev. Genet. 17, 333–351.
Gough, E., Shaikh, H., and Manges, A.R. (2011). Systematic Review of Intestinal Microbiota Transplantation
(Fecal Bacteriotherapy) for Recurrent Clostridium difficile Infection. Clin. Infect. Dis. 53, 994–1002.
Grice, E.A. (2014). The skin microbiome: potential for novel diagnostic and therapeutic approaches to
cutaneous disease. Semin. Cutan. Med. Surg. 33, 98–103.
Grice, E.A. (2015). The intersection of microbiome and host at the skin interface : genomic- and
metagenomic-based insights. Genome Res. 25, 1514–1520.
Grice, E.A., and Segre, J.A. (2011). The skin microbiome. Nat. Rev. Microbiol. 9, 244–253.
Grice, E.A., and Segre, J.A. (2012). The Human Microbiome: Our Second Genome. Annu. Rev. Genomics
Hum. Genet. 13, 151–170.
159
Grice, E.A., Kong, H.H., Conlan, S., Deming, C.B., Davis, J., Young, A.C., Bouffard, G.G., Blakesley, R.W.,
Murray, P.R., Green, E.D., et al. (2009a). Topographical and temporal diversity of the human skin
microbiome. Science 324, 1190–1192.
Grice, E.A., Kong, H.H., Conlan, S., Deming, C.B., Davis, J., Young, A.C., NISC Comparative Sequencing
Program, Bouffard, G.G., Blakesley, R.W., Murray, P.R., et al. (2009b). Topographical and temporal diversity
of the human skin. Science 324, 1190–1192.
Grice, E.A., Kong, H.H., Renaud, G., Young, A.C., NISC Comparative Sequencing Program, Bouffard,
G.G., Blakesley, R.W., Wolfsberg, T.G., Turner, M.L., and Segre, J.A. (2008). A diversity profile of the human
skin microbiota. Genome Res. 18, 1043–1050.
Guéniche, A., Bastien, P., Ovigne, J.M., Kermici, M., Courchay, G., Chevalier, V., Breton, L., and Castiel-
Higounenc, I. (2009). Bifidobacterium longum lysate, a new ingredient for reactive skin. Exp. Dermatol. 19,
e1–e8.
Haas, B.J., Gevers, D., Earl, A.M., Feldgarden, M., Ward, D. V, Giannoukos, G., Ciulla, D., Tabbaa, D.,
Highlander, S.K., Desantis, T.Z., et al. (2011). Chimeric 16S rRNA sequence formation and detection in
Sanger and 454-pyrosequenced PCR amplicons. Genome Res. 21, 494–504.
Halwachs, B., Madhusudhan, N., Krause, R., Nilsson, R.H., Moissl-Eichinger, C., Högenauer, C., Thallinger,
G.G., and Gorkiewicz, G. (2017). Critical Issues in Mycobiota Analysis. Front. Microbiol. 8, 180.
Hammer, T.J., Janzen, D.H., Hallwachs, W., Jaffe, S.L., and Fierer, N. (2017). Caterpillars lack a resident gut
microbiome. bioRxiv.
Hampton-Marcell, J.T., Lopez, J. V., and Gilbert, J.A. (2017). The human microbiome: an emerging tool in
forensics. Microb. Biotechnol. 10, 228–230.
Handley, S.A. (2016). The virome: a missing component of biological interaction networks in health and
disease. Genome Med. 8, 32.
Hannigan, G.D., Hodkinson, B.P., McGinnis, K., Tyldsley, A.S., Anari, J.B., Horan, A.D., Grice, E.A., and
Mehta, S. (2014). Culture-independent pilot study of microbiota colonizing open fractures and association
with severity, mechanism, location, and complication from presentation to early outpatient follow-up. J.
Orthop. Res. 32, 597–605.
Hannigan, G.D., Meisel, J.S., Tyldsley, A.S., Zheng, Q., Hodkinson, B.P., Sanmiguel, A.J., Minot, S.,
Bushman, F.D., and Grice, A. (2015). The Human Skin Double-Stranded DNA Virome : Topographical and
Temporal Diversity , Genetic Enrichment , and Dynamic Associations with the Host Microbiome. 6, 1–13.
Hanski, I., von Hertzen, L., Fyhrquist, N., Koskinen, K., Torppa, K., Laatikainen, T., Karisola, P., Auvinen,
P., Paulin, L., Makela, M.J., et al. (2012). Environmental biodiversity, human microbiota, and allergy are
interrelated. Proc. Natl. Acad. Sci. 109, 8334–8339.
Harris, E.E., and Meyer, D. (2006). The molecular signature of selection underlying human adaptations. Am.
J. Phys. Anthropol. 131, 89–130.
He, Y., Caporaso, J.G., Jiang, X.-T., Sheng, H.-F., Huse, S.M., Rideout, J.R., Edgar, R.C., Kopylova, E.,
Walters, W.A., Knight, R., et al. (2015). Stability of operational taxonomic units: an important but neglected
property for analyzing microbial diversity. Microbiome 3, 20.
Heather, J.M., and Chain, B. (2016). The sequence of sequencers: The history of sequencing DNA. Genomics
107, 1–8.
Hill, J.E., Penny, S.L., Crowell, K.G., Goh, S.H., and Hemmingsen, S.M. (2004). cpnDB: a chaperonin
sequence database. Genome Res. 14, 1669–1675.
160
Hilmi, D., Parcina, M., Stollewerk, D., Ostrop, J., Josten, M., Meilaender, A., Zaehringer, U., Wichelhaus,
T.A., Bierbaum, G., Heeg, K., et al. (2014). Heterogeneity of Host TLR2 Stimulation by Staphylocoocus
aureus Isolates. PLoS One 9, e96416.
Hoang, B.X., Shaw, G., Pham, P., and Levine, S.A. (2010). Lactobacillus rhamnosus cell lysate in the
management of resistant childhood atopic eczema. Inflamm. Allergy Drug Targets 9, 192–196.
Holmes, C.J., Plichta, J.K., Gamelli, R.L., and Radek, K.A. (2015). Dynamic Role of Host Stress Responses in
Modulating the Cutaneous Microbiome: Implications for Wound Healing and Infection. Adv. Wound Care 4,
24–37.
Hosein, S., Rodríguez-Cortés, A., Blake, D.P., Allenspach, K., Alberola, J., and Solano-Gallego, L. (2015).
Transcription of Toll-Like Receptors 2, 3, 4 and 9, FoxP3 and Th17 Cytokines in a Susceptible Experimental
Model of Canine Leishmania infantum Infection. PLoS One 10, e0140325.
Hospodsky, D., Pickering, A.J., Julian, T.R., Miller, D., Gorthala, S., Boehm, A.B., and Peccia, J. (2014). Hand
bacterial communities vary across two different human populations. Microbiology 160, 1144–1152.
Hughes, J.B., and Hellmann, J.J. (2005). The Application of Rarefaction Techniques to Molecular Inventories
of Microbial Diversity. In Methods in Enzymology 397, 292-308
Huizenga, T., Kado, J., Mehregan, D.R., and Diamond, S. (2015). Identifying Toll-like Receptor Expression
in Cutaneous Sarcoidosis. Am. J. Dermatopathol. 37, 67–72.
The Human Microbiome Project Consortium. (2012a). A framework for human microbiome research.
Nature 486, 215–221.
The Human Microbiome Project Consortium. (2012b). Structure, function and diversity of the healthy
human microbiome. Nature 486, 207–214.
Huseini, H.F., Rahimzadeh, G., Fazeli, M.R., Mehrazma, M., Salehi, M., Hooman, Y., and al., et (2012).
Evaluation of wound healing activities of kefir products. Burns 38, 719–723.
Ip, C.L.C., Loose, M., Tyson, J.R., de Cesare, M., Brown, B.L., Jain, M., Leggett, R.M., Eccles, D.A., Zalunin,
V., Urban, J.M., et al. (2015). MinION Analysis and Reference Consortium: Phase 1 data release and analysis.
F1000Research 4, 1075.
Janardhanam, S.B., Prakasam, S., Swaminathan, V.T., Kodumudi, K.N., Zunt, S.L., and Srinivasan, M. (2012).
Differential expression of TLR-2 and TLR-4 in the epithelial cells in oral lichen planus. Arch. Oral Biol. 57,
495–502.
Janda, J.M., and Abbott, S.L. (2007). 16S rRNA gene sequencing for bacterial identification in the diagnostic
laboratory: pluses, perils, and pitfalls. J. Clin. Microbiol. 45, 2761–2764.
Jo, J.H., Deming, C., Kennedy, E.A., Conlan, S., Polley, E.C., Ng, W., Segre, J.A., Kong, H.H., and Kong,
H.H. (2016). Diverse Human Skin Fungal Communities in Children Converge in Adulthood. J. Invest.
Dermatol. 136, 2356–2363.
Jumpstart Consortium Human Microbiome Project Data Generation Working Group (2012). Evaluation of
16S rDNA-based community profiling for human microbiome research. PLoS One 7, e39315.
Jun, S.-R., Robeson, M.S., Hauser, L.J., Schadt, C.W., and Gorin, A.A. (2015). PanFP: pangenome-based
functional profiles for microbial communities. BMC Res Notes 8, 479.
Kang, J.Y., and Lee, J.O. (2011). Structural Biology of the Toll-Like Receptor Family. Annu. Rev. Biochem.
80, 917–941.
161
Karst, S.M., Dueholm, M.S., McIlroy, S.J., Kirkegaard, R.H., Nielsen, P.H., and Albertsen, M. (2016).
Thousands of primer-free, high-quality, full-length SSU rRNA sequences from all domains of life. bioRxiv.
Kasperkovitz, P.V., Khan, N.S., Tam, J.M., Mansour, M.K., Davids, P.J., and Vyas, J.M. (2011). Toll-Like
Receptor 9 Modulates Macrophage Antifungal Effector Function during Innate Recognition of Candida
albicans and Saccharomyces cerevisiae. Infect. Immun. 79, 4858–4867.
Kathrani, A., House, A., Catchpole, B., Murphy, A., Werling, D., and Allenspach, K. (2011). Breed-
independent toll-like receptor 5 polymorphisms show association with canine inflammatory bowel disease.
Tissue Antigens 78, 94–101.
Kathrani, A., Holder, A., Catchpole, B., Alvarez, L., Simpson, K., Werling, D., and Allenspach, K. (2012).
TLR5 risk-associated haplotype for canine inflammatory bowel disease confers hyper-responsiveness to
flagellin. PLoS One 7, e30117.
Kathrani, A., House, A., Catchpole, B., Murphy, A., German, A., Werling, D., and Allenspach, K. (2010).
Polymorphisms in the TLR4 and TLR5 gene are significantly associated with inflammatory bowel disease in
German shepherd dogs. PLoS One 5, e15740.
Kawai, T., and Akira, S. (2009). The roles of TLRs, RLRs and NLRs in pathogen recognition. Int. Immunol.
21, 317–337.
Kawai, T., and Akira, S. (2010). The role of pattern-recognition receptors in innate immunity: update on Toll-
like receptors. Nat. Immunol. 11, 373–384.
Kelly, B.J., Gross, R., Bittinger, K., Sherrill-Mix, S., Lewis, J.D., Collman, R.G., Bushman, F.D., and Li, H.
(2015). Power and sample-size estimation for microbiome studies using pairwise distances and
PERMANOVA. Bioinformatics 31, 2461–2468.
Kennedy, E.A., Connolly, J., Hourihane, J.O., Fallon, P.G., McLean, W.H.I., Murray, D., Jo, J.H., Segre, J.A.,
Kong, H.H., and Irvine, A.D. (2017). Skin microbiome before development of atopic dermatitis: Early
colonization with commensal staphylococci at 2 months is associated with a lower risk of atopic dermatitis at
1 year. J. Allergy Clin. Immunol. 139, 166–172.
Kettleson, E.M., Adhikari, A., Vesper, S., Coombs, K., Indugula, R., and Reponen, T. (2015). Key
determinants of the fungal and bacterial microbiomes in homes. Environ. Res. 138, 130–135.
Kim, J., Ochoa, M.-T., Krutzik, S.R., Takeuchi, O., Uematsu, S., Legaspi, A.J., Brightbill, H.D., Holland, D.,
Cunliffe, W.J., Akira, S., et al. (2002). Activation of toll-like receptor 2 in acne triggers inflammatory cytokine
responses. J. Immunol. 169, 1535–1541.
Klappenbach, J.A., Saxman, P.R., Cole, J.R., and Schmidt, T.M. (2001). rrndb: the Ribosomal RNA Operon
Copy Number Database. Nucleic Acids Res. 29, 181–184.
Klindworth, A., Pruesse, E., Schweer, T., Peplies, J., Quast, C., Horn, M., and Glockner, F.O. (2013).
Evaluation of general 16S ribosomal RNA gene PCR primers for classical and next-generation sequencing-
based diversity studies. Nucleic Acids Res. 41, e1–e1.
Knights, D., Silverberg, M.S., Weersma, R.K., Gevers, D., Dijkstra, G., Huang, H., Tyler, A.D., van
Sommeren, S., Imhann, F., Stempak, J.M., et al. (2014). Complex host genetics influence the microbiome in
inflammatory bowel disease. Genome Med. 6, 107.
Kong, H.H., Andersson, B., Clavel, T., Common, J.E., Jackson, S.A., Olson, N.D., Segre, J.A., and Traidl-
Hoffmann, C. (2017). Performing Skin Microbiome Research: A Method to the Madness. J. Invest. Dermatol.
137, 561–568.
162
Kong, H.H., Oh, J., Deming, C., Conlan, S., Grice, E.A., Beatson, M.A., Nomicos, E., Polley, E.C.,
Komarow, H.D., Murray, P.R., et al. (2012). Temporal shifts in the skin microbiome associated with disease
flares and treatment in children with atopic dermatitis. Genome Res. 22, 850–859.
Kopylova, E., Navas-Molina, J.A., Mercier, C., Xu, Z.Z., Mahé, F., He, Y., Zhou, H. W., Rognes, T.,
Caporaso, J.G., and Knight, R. (2016). Open-Source Sequence Clustering Methods Improve the State Of the
Art. mSystems 1, e00003-15.
Koren, S., Walenz, B.P., Berlin, K., Miller, J.R., Bergman, N.H., and Phillippy, A.M. (2017). Canu: scalable
and accurate long-read assembly via adaptive k -mer weighting and repeat separation. Genome Res. 27, 722–
736.
Kruskal, W.H., and Wallis, W.A. (1952). Use of Ranks in One-Criterion Variance Analysis. J. Am. Stat. Assoc.
47, 583.
Krutmann, J. (2009). Pre- and probiotics for human skin. J. Dermatol. Sci. 54, 1–5.
Krutzik, S.R., Ochoa, M.T., Sieling, P.A., Uematsu, S., Ng, Y.W., Legaspi, A., Liu, P.T., Cole, S.T., Godowski,
P.J., Maeda, Y., et al. (2003). Activation and regulation of Toll-like receptors 2 and 1 in human leprosy. Nat.
Med. 9, 525–532.
Kubinak, J.L., and Round, J.L. (2012). Toll-like receptors promote mutually beneficial commensal-host
interactions. PLoS Pathog. 8, 1.
Kuczynski, J., Lauber, C.L., Walters, W. a., Parfrey, L.W., Clemente, J.C., Gevers, D., and Knight, R. (2011).
Experimental and analytical tools for studying the human microbiome. Nat. Rev. Genet. 13, 47–58.
Kumar, H., Kawai, T., and Akira, S. (2009). Pathogen recognition in the innate immune response. Biochem. J.
420, 1–16.
Kumar, H., Kawai, T., and Akira, S. (2011). Pathogen recognition by the innate immune system. Int. Rev.
Immunol. 30, 16–34.
Kumar, S., Ingle, H., Prasad, D.V.R., and Kumar, H. (2013). Recognition of bacterial infection by innate
immune sensors. Crit. Rev. Microbiol. 39, 229–246.
Kumar, V., and Sharma, A. (2010). Mast cells: Emerging sentinel innate immune cells with diverse role in
immunity. Mol. Immunol. 48, 14–25.
Kuroki, K., Stoker, A.M., Sims, H.J., and Cook, J.L. (2010). Expression of Toll-like receptors 2 and 4 in stifle
joint synovial tissues of dogs with or without osteoarthritis. Am. J. Vet. Res. 71, 750–754.
La Rosa, P.S., Brooks, J.P., Deych, E., Boone, E.L., Edwards, D.J., Wang, Q., Sodergren, E., Weinstock, G.,
and Shannon, W.D. (2012). Hypothesis Testing and Power Calculations for Taxonomic-Based Human
Microbiome Data. PLoS One 7, e52078.
Lai, Y., Di Nardo, A., Nakatsuji, T., Leichtle, A., Yang, Y., Cogen, A.L., Wu, Z.-R., Hooper, L. V, Schmidt,
R.R., von Aulock, S., et al. (2009). Commensal bacteria regulate Toll-like receptor 3-dependent inflammation
after skin injury. Nat. Med. 15, 1377–1382.
Langille, M.G.I., Zaneveld, J., Caporaso, J.G., McDonald, D., Knights, D., Reyes, J. a, Clemente, J.C.,
Burkepile, D.E., Vega Thurber, R.L., Knight, R., et al. (2013). Predictive functional profiling of microbial
communities using 16S rRNA marker gene sequences. Nat. Biotechnol. 31, 814–821.
Lauber, C.L., Zhou, N., Gordon, J.I., Knight, R., and Fierer, N. (2010). Effect of storage conditions on the
assessment of bacterial community structure in soil and human-associated samples. FEMS Microbiol. Lett.
307, 80–86.
163
Lax, S., Hampton-Marcell, J.T., Gibbons, S.M., Colares, G.B., Smith, D., Eisen, J.A., and Gilbert, J.A. (2015).
Forensic analysis of the microbiome of phones and shoes. Microbiome 3, 21.
Leber, A., Viladomiu, M., Hontecillas, R., Abedi, V., Philipson, C., Hoops, S., Howard, B., and Bassaganya-
Riera, J. (2015). Systems Modeling of Interactions between Mucosal Immunity and the Gut Microbiome
during Clostridium difficile Infection. PLoS One 10, e0134849.
Lehtimäki, J., Karkman, A., Laatikainen, T., Paalanen, L., von Hertzen, L., Haahtela, T., Hanski, I., and
Ruokolainen, L. (2017). Patterns in the skin microbiota differ in children and teenagers between rural and
urban environments. Sci. Rep. 7, 45651.
Leifer, C.A., McConkey, C., Li, S., Chassaing, B., Gewirtz, A.T., and Ley, R.E. (2014). Linking genetic
variation in human Toll-like receptor 5 genes to the gut microbiome’s potential to cause inflammation.
Immunol. Lett. 162, 3–9.
Leigh, I.M., and Watt, F.M. (1994). The keratinocyte handbook (New York: Cambridge University Press).
Lesiak, A., Smolewski, P., Sobolewska-Sztychny, D., Sysa-Jedrzejowska, A., and Narbutt, J. (2012). The Role
of T-Regulatory Cells and Toll-Like Receptors 2 and 4 In Atopic Dermatitis. Scand. J. Immunol. 76, 405–410.
Leung, M.H.Y., Wilkins, D., and Lee, P.K.H. (2015). Insights into the pan-microbiome: skin microbial
communities of Chinese individuals differ from other racial groups. Sci. Rep. 5, 11845.
Levene, M.J., Korlach, J., Turner, S.W., Foquet, M., Craighead, H.G., and Webb, W.W. (2003). Zero-Mode
Waveguides for Single-Molecule Analysis at High Concentrations. Science 299, 682–686.
Lew, L.C., and Liong, M.T. (2013). Bioactives from probiotics for dermal health: functions and benefits. J.
Appl. Microbiol. 114, 1241–1253.
Li, C., Chng, K.R., Boey, E.J.H., Ng, A.H.Q., Wilm, A., and Nagarajan, N. (2016). INC-Seq: accurate single
molecule reads using nanopore sequencing. Gigascience 5, 34.
Lindblad-Toh, K., Wade, C.M., Mikkelsen, T.S., Karlsson, E.K., Jaffe, D.B., Kamal, M., Clamp, M., Chang,
J.L., Kulbokas, E.J., Zody, M.C., et al. (2005). Genome sequence, comparative analysis and haplotype
structure of the domestic dog. Nature 438, 803–819.
Links, M.G., Dumonceaux, T.J., Hemmingsen, S.M., Hill, J.E., and Dalin, E. (2012). The Chaperonin-60
Universal Target Is a Barcode for Bacteria That Enables De Novo Assembly of Metagenomic Sequence Data.
PLoS One 7, e49755.
Loman, N.J., Misra, R. V, Dallman, T.J., Constantinidou, C., Gharbia, S.E., Wain, J., and Pallen, M.J. (2012).
Performance comparison of benchtop high-throughput sequencing platforms. Nat. Biotechnol. 30, 434–439.
Loman, N.J., Quick, J., and Simpson, J.T. (2015). A complete bacterial genome assembled de novo using only
nanopore sequencing data. Nat. Methods 12, 733–735.
Lozupone, C., and Knight, R. (2005). UniFrac : a New Phylogenetic Method for Comparing Microbial
Communities. Appl. Environ. Microbiol. 71, 8228–8235.
Lozupone, C., Lladser, M.E., Knights, D., Stombaugh, J., and Knight, R. (2011). UniFrac: an effective
distance metric for microbial community comparison. ISME J. 5, 169–172.
Lozupone, C.A., and Knight, R. (2007). Global patterns in bacterial diversity. Proc. Natl. Acad. Sci. U. S. A.
104, 11436–11440.
Lozupone, C.A., and Knight, R. (2008). Species divergence and the measurement of microbial diversity.
FEMS Microbiol. Rev. 32, 557–578.
164
Lozupone, C.A., Hamady, M., Kelley, S.T., and Knight, R. (2007). Quantitative and Qualitative Diversity
Measures Lead to Different Insights into Factors That Structure Microbial Communities. Appl. Environ.
Microbiol. 73, 1576–1585.
Lund, P.A. (2009). Multiple chaperonins in bacteria – why so many? FEMS Microbiol. Rev. 33, 785–800.
Luo, C., Rodriguez-R, L.M., and Konstantinidis, K.T. (2013). A User’s Guide to Quantitative and
Comparative Analysis of Metagenomic Datasets. In Methods in Enzymology, pp. 525–547.
Mao, D.P., Zhou, Q., Chen, C.-Y., and Quan, Z.X. (2012). Coverage evaluation of universal bacterial primers
using the metagenomic datasets. BMC Microbiol. 12, 66.
Marchesi, J.R., and Ravel, J. (2015). The vocabulary of microbiome research: a proposal. Microbiome 3, 31.
Mardis, E.R. (2017). DNA sequencing technologies: 2006-2016. Nat. Protoc. 12, 213–218.
Matin, N., Tabatabaie, O., Mohammadinejad, P., and Rezaei, N. (2015). Therapeutic targeting of Toll-like
receptors in cutaneous disorders. Expert Opin. Ther. Targets 19, 1651–1663.
McDonald, D., Price, M.N., Goodrich, J., Nawrocki, E.P., DeSantis, T.Z., Probst, A., Andersen, G.L.,
Knight, R., and Hugenholtz, P. (2012). An improved Greengenes taxonomy with explicit ranks for ecological
and evolutionary analyses of bacteria and archaea. ISME J. 6, 610–618.
McMahon, L.A., House, A.K., Catchpole, B., Elson-Riggins, J., Riddle, A., Smith, K., Werling, D., Burgener,
I.A., and Allenspach, K. (2010). Expression of Toll-like receptor 2 in duodenal biopsies from dogs with
inflammatory bowel disease is associated with severity of disease. Vet. Immunol. Immunopathol. 135, 158–
163.
Meadow, J.F., Altrichter, A.E., and Green, J.L. (2014). Mobile phones carry the personal microbiome of their
owners. PeerJ 2, e447.
Meason-Smith, C., Diesel, A., Patterson, A.P., Older, C.E., Johnson, T.J., Mansell, J.M., Suchodolski, J.S., and
Rodrigues Hoffmann, A. (2017). Characterization of the cutaneous mycobiota in healthy and allergic cats
using next generation sequencing. Vet. Dermatol. 28, 71–e17.
Meason-smith, C., Diesel, A., Patterson, A.P., Older, C.E., Mansell, J.M., Suchodolski, J.S., and Hoffmann,
A.R. (2015). What is living on your dog ’s skin ? Characterization of the canine cutaneous mycobiota and
fungal dysbiosis in canine allergic dermatitis. FEMS Microbiol. Ecol. 91, 1–12.
Meisel, J.S., Hannigan, G.D., Tyldsley, A.S., Sanmiguel, A.J., Hodkinson, B.P., Zheng, Q., and Grice, E.A.
(2016). Skin Microbiome Surveys Are Strongly Influenced by Experimental Design. J. Invest. Dermatol. 136,
947–956.
Melo, G.D., Silva, J.E.S., Grano, F.G., Homem, C.G., and Machado, G.F. (2014). Compartmentalized gene
expression of toll-like receptors 2, 4 and 9 in the brain and peripheral lymphoid organs during canine visceral
leishmaniasis. Parasite Immunol. 36, 726–731.
Mercier, E., Peters, I.R., Day, M.J., Clercx, C., and Peeters, D. (2012). Toll- and NOD-like receptor mRNA
expression in canine sino-nasal aspergillosis and idiopathic lymphoplasmacytic rhinitis. Vet. Immunol.
Immunopathol. 145, 618–624.
Miletto, M., and Lindow, S.E. (2015). Relative and contextual contribution of different sources to the
composition and abundance of indoor air bacteria in residences. Microbiome 3, 61.
Miller, W.H., Griffin, C.E., and Campbell, K.L. (2013). Muller & Kirk’s Small Animal Dermatology (St Louis:
Saunders).
165
Misic, A.M., Davis, M.F., Tyldsley, A.S., Hodkinson, B.P., Tolomeo, P., Hu, B., Nachamkin, I., Lautenbach,
E., Morris, D.O., and Grice, E.A. (2015). The shared microbiota of humans and companion animals as
evaluated from Staphylococcus carriage sites. Microbiome 3, 2.
Mizrahi-Man, O., Davenport, E.R., and Gilad, Y. (2013). Taxonomic Classification of Bacterial 16S rRNA
Genes Using Short Sequencing Reads: Evaluation of Effective Study Designs. PLoS One 8, 18–23.
Morey, M., Fernández-Marmiesse, A., Castiñeiras, D., Fraga, J.M., Couce, M.L., and Cocho, J.A. (2013). A
glimpse into past, present, and future DNA sequencing. Mol. Genet. Metab. 110, 3–24.
Muehleisen, B., Jiang, S.B., Gladsjo, J.A., Gerber, M., Hata, T., and Gallo, R.L. (2012). Distinct Innate
Immune Gene Expression Profiles in Non-Melanoma Skin Cancer of Immunocompetent and
Immunosuppressed Patients. PLoS One 7, e40754.
Mukherjee, S., Ganguli, D., and Majumder, P.P. (2014). Global Footprints of Purifying Selection on Toll-Like
Receptor Genes Primarily Associated with Response to Bacterial Infections in Humans. Genome Biol. Evol.
6, 551–558.
Mukherjee, S., Sarkar-Roy, N., Wagener, D.K., and Majumder, P.P. (2009). Signatures of natural selection are
not uniform across genes of innate immune system, but purifying selection is the dominant signature. Proc.
Natl. Acad. Sci. 106, 7073–7078.
Myles, I.A., Williams, K.W., Reckhow, J.D., Jammeh, M.L., Pincus, N.B., Sastalla, I., Saleem, D., Stone, K.D.,
and Datta, S.K. (2016). Transplantation of human skin microbiota in models of atopic dermatitis. JCI Insight
1, e86955.
Naik, S., Bouladoux, N., Linehan, J.L., Han, S.-J., Harrison, O.J., Wilhelm, C., Conlan, S., Himmelfarb, S.,
Byrd, A.L., Deming, C., et al. (2015). Commensal-dendritic-cell interaction specifies a unique protective skin
immune signature. Nature 520, 104–108.
Naik, S., Bouladoux, N., Wilhelm, C., Molloy, M.J., Salcedo, R., Kastenmuller, W., Deming, C., Quinones, M.,
Koo, L., Conlan, S., et al. (2012). Compartmentalized Control of Skin Immunity by Resident Commensals.
Science 337, 1115–1119.
Nakamizo, S., Egawa, G., Honda, T., Nakajima, S., Belkaid, Y., and Kabashima, K. (2015). Commensal
bacteria and cutaneous immunity. Semin. Immunopathol. 37, 73–80.
Nakatsuji, T., Chiang, H.-I., Jiang, S.B., Nagarajan, H., Zengler, K., and Gallo, R.L. (2013). The microbiome
extends to subepidermal compartments of normal skin. Nat. Commun. 4, 1431.
Nestle, F.O., Di Meglio, P., Qin, J.-Z., and Nickoloff, B.J. (2009). Skin immune sentinels in health and
disease. Nat. Rev. Immunol. 9, 679–691.
Netea, M.G., Wijmenga, C., and O’Neill, L. a J. (2012). Genetic variation in Toll-like receptors and disease
susceptibility. Nat. Immunol. 13, 535–542.
O’Neill, L.A.J., Golenbock, D., and Bowie, A.G. (2013). The history of Toll-like receptors — redefining
innate immunity. Nat. Rev. Immunol. 13, 453–460.
Oh, J., Byrd, A.L., Deming, C., Conlan, S., Barnabas, B., Blakesley, R., Bouffard, G., Brooks, S., Coleman, H.,
Dekhtyar, M., et al. (2014). Biogeography and individuality shape function in the human skin metagenome.
Nature 514, 59–64.
Oh, J., Byrd, A.L., Park, M., Kong, H.H., and Segre, J.A. (2016). Temporal Stability of the Human Skin
Microbiome. Cell 165, 854–866.
166
Oh, J., Conlan, S., Polley, E., Segre, J. a, and Kong, H.H. (2012). Shifts in human skin and nares microbiota
of healthy children and adults. Genome Med. 4, 77.
Oh, J., Freeman, A.F., Park, M., Sokolic, R., Candotti, F., Holland, S.M., Segre, J.A., and Kong, H.H. (2013).
The altered landscape of the human skin microbiome in patients with primary immunodeficiencies. Genome
Res. 23, 2103–2114.
Olasagasti, F., Lieberman, K.R., Benner, S., Cherf, G.M., Dahl, J.M., Deamer, D.W., and Akeson, M. (2010).
Replication of individual DNA molecules under electronic control using a protein nanopore. Nat.
Nanotechnol. 5, 798–806.
Olivry, T., and Bizikova, P. (2013). A systematic review of randomized controlled trials for prevention or
treatment of atopic dermatitis in dogs: 2008-2011 update. Vet. Dermatol. 24, 97–e26.
Pasparakis, M., Haase, I., and Nestle, F.O. (2014). Mechanisms regulating skin immunity and inflammation.
Nat Rev Immunol 14, 289–301.
Patin, N. V., Kunin, V., Lidström, U., and Ashby, M.N. (2013). Effects of OTU Clustering and PCR Artifacts
on Microbial Diversity Estimates. Microb. Ecol. 65, 709–719.
Paulino, L.C., Tseng, C. H., Strober, B.E., and Blaser, M.J. (2006). Molecular Analysis of Fungal Microbiota in
Samples from Healthy Human Skin and Psoriatic Lesions. J. Clin. Microbiol. 44, 2933–2941.
Peleg, A.Y., Hogan, D.A., and Mylonakis, E. (2010). Medically important bacterial–fungal interactions. Nat.
Rev. Microbiol. 8, 340–349.
Perez Perez, G.I., Gao, Z., Jourdain, R., Ramirez, J., Gany, F., Clavaud, C., Demaude, J., Breton, L., and
Blaser, M.J. (2016). Body Site Is a More Determinant Factor than Human Population Diversity in the Healthy
Skin Microbiome. PLoS One 11, e0151990.
Pierezan, F., Olivry, T., Paps, J.S., Lawhon, S.D., Wu, J., Steiner, J.M., Suchodolski, J.S., and Hoffmann, A.R.
(2016). The skin microbiome in allergen-induced canine atopic dermatitis. Vet Dermatol. 2016 27, 332-e82
Potaczek, D., Nastalek, M., Okumura, K., Wojas-Pelc, A., Undas, A., and Nishiyama, C. (2011). An
association of TLR2-16934A>T polymorphism and severity/phenotype of atopic dermatitis. J. Eur. Acad.
Dermatology Venereol. 25, 715–721.
Price, L.B., Liu, C.M., Melendez, J.H., Frankel, Y.M., Engelthaler, D., Aziz, M., Bowers, J., Rattray, R., Ravel,
J., Kingsley, C., et al. (2009). Community analysis of chronic wound bacteria using 16S rRNA gene-based
pyrosequencing: impact of diabetes and antibiotics on chronic wound microbiota. PLoS One 4, e6462.
Quach, H., Wilson, D., Laval, G., Patin, E., Manry, J., Guibert, J., Barreiro, L.B., Nerrienet, E., Verschoor, E.,
Gessain, A., et al. (2013). Different selective pressures shape the evolution of Toll-like receptors in human
and African great ape populations. Hum. Mol. Genet. 22, 4829–4840.
Quast, C., Pruesse, E., Yilmaz, P., Gerken, J., Schweer, T., Yarza, P., Peplies, J., and Glöckner, F.O. (2013).
The SILVA ribosomal RNA gene database project: improved data processing and web-based tools. Nucleic
Acids Res. 41, D590–D596.
Quick, J., and Loman, N.J. Thar she blows! Ultra-long read methods for nanopore sequencing. Oral
presentation at London Calling 2017 by Oxford Nanopore Technologies.
Quintana-Murci, L. (2016). Understanding rare and common diseases in the context of human evolution.
Genome Biol. 17, 225.
Rajendhran, J., and Gunasekaran, P. (2011). Microbial phylogeny and diversity: Small subunit ribosomal RNA
sequence analysis and beyond. Microbiol. Res. 166, 99–110.
167
Reid, G., Younes, J.A., Van der Mei, H.C., Gloor, G.B., Knight, R., and Busscher, H.J. (2011). Microbiota
restoration: natural and supplemented recovery of human microbial communities. Nat. Rev. Microbiol. 9, 27–
38.
Rensburg, J.J. Van, Lin, H., Gao, X., Toh, E., Fortney, K.R., Ellinger, S., Zwickl, B., Janowicz, D.M., Katz,
B.P., Nelson, D.E., et al. (2015). The Human Skin Microbiome Associates with the Outcome of and Is
Influenced by Bacterial Infection. 6, 1–13.
Reuter, J.A., Spacek, D. V, and Snyder, M.P. (2015). High-throughput sequencing technologies. Mol. Cell 58,
586–597.
Richter, M., and Rossello-Mora, R. (2009). Shifting the genomic gold standard for the prokaryotic species
definition. Proc. Natl. Acad. Sci. 106, 19126–19131.
Rideout, J.R., He, Y., Navas-Molina, J.A., Walters, W.A., Ursell, L.K., Gibbons, S.M., Chase, J., McDonald,
D., Gonzalez, A., Robbins-Pianka, A., et al. (2014). Subsampled open-reference clustering creates consistent,
comprehensive OTU definitions and scales to billions of sequences. PeerJ 2, e545.
Riggio, M.P., Lappin, D.F., and Bennett, D. (2014). Bacteria and Toll-like receptor and cytokine mRNA
expression profiles associated with canine arthritis. Vet. Immunol. Immunopathol. 160, 158–166.
Roach, J.C., Glusman, G., Rowen, L., Kaur, A., Purcell, M.K., Smith, K.D., Hood, L.E., and Aderem, A.
(2005). The evolution of vertebrate Toll-like receptors. Proc. Natl. Acad. Sci. U. S. A. 102, 9577–9582.
Rodrigues Hoffmann, A., Patterson, A.P., Diesel, A., Lawhon, S.D., Ly, H.J., Elkins Stephenson, C., Mansell,
J., Steiner, J.M., Dowd, S.E., Olivry, T., et al. (2014). The skin microbiome in healthy and allergic dogs. PLoS
One 9, e83197.
Rogers, G.B., and Bruce, K.D. (2010). Next-Generation Sequencing in the Analysis of Human Microbiota.
Mol. Diagn. Ther. 14, 343–350.
Rognes, T., Flouri, T., Nichols, B., Quince, C., and Mahé, F. (2016). VSEARCH: a versatile open source tool
for metagenomics. PeerJ 4, e2584.
Rossi-Tamisier, M., Benamar, S., Raoult, D., and Fournier, P.-E. (2015). Cautionary tale of using 16S rRNA
gene sequence similarity values in identification of human-associated bacterial species. Int. J. Syst. Evol.
Microbiol. 65, 1929–1934.
Rothberg, J.M., Hinz, W., Rearick, T.M., Schultz, J., Mileski, W., Davey, M., Leamon, J.H., Johnson, K.,
Milgrew, M.J., Edwards, M., et al. (2011). An integrated semiconductor device enabling non-optical genome
sequencing. Nature 475, 348–352.
Ruedisueli, F.L., Eastwood, N.J., Gunn, N.K., and Watson, T.D.G. (1998). The measurement of skin pH in
normal dogs of different breeds. In Advances in Veterinary Dermatology III, K. Kwochka, T. Willemse, and
C. Von Tscharner, eds. (Boston: Butterworth-Heinemann Ltd), p. 596.
Salem, S.A.M., Abu-Zeid, R.M., and Nada, O.H. (2013). Immunohistochemical study of toll-like receptors 1
and 2 expression in cutaneous lichen planus lesions. Arch. Dermatol. Res. 305, 125–131.
Salter, S.J., Cox, M.J., Turek, E.M., Calus, S.T., Cookson, W.O., Moffatt, M.F., Turner, P., Parkhill, J., Loman,
N.J., and Walker, A.W. (2014). Reagent and laboratory contamination can critically impact sequence-based
microbiome analyses. BMC Biol. 12, 87
Sanford, J. a., and Gallo, R.L. (2013). Functions of the skin microbiota in health and disease. Semin.
Immunol. 25, 370–377.
168
Sanmiguel, A., and Grice, E.A. (2015). Interactions between host factors and the skin microbiome. Cell. Mol.
Life Sci. 72, 1499–1515.
Santiago, A., Panda, S., Mengels, G., Martinez, X., Azpiroz, F., Dore, J., Guarner, F., and Manichanh, C.
(2014). Processing faecal samples: a step forward for standards in microbial community analysis. BMC
Microbiol. 14, 112.
Schloss, P.D., and Handelsman, J. (2004). Status of the microbial census. Microbiol. Mol. Biol. Rev. 68, 686–
691.
Schloss, P.D., Gevers, D., and Westcott, S.L. (2011). Reducing the effects of PCR amplification and
sequencing artifacts on 16S rRNA-based studies. PLoS One 6, e27310.
Schloss, P.D., Girard, R.A., Martin, T., Edwards, J., and Thrash, J.C. (2016b). Status of the Archaeal and
Bacterial Census: an Update. MBio 7, e00201-16
Schloss, P.D., Jenior, M.L., Koumpouras, C.C., Westcott, S.L., and Highlander, S.K. (2016a). Sequencing 16S
rRNA gene fragments using the PacBio SMRT DNA sequencing system. PeerJ 4, e1869.
Schloss, P.D., Westcott, S.L., Ryabin, T., Hall, J.R., Hartmann, M., Hollister, E.B., Lesniewski, R.A., Oakley,
B.B., Parks, D.H., Robinson, C.J., et al. (2009). Introducing mothur: Open-Source, Platform-Independent,
Community-Supported Software for Describing and Comparing Microbial Communities. Appl. Environ.
Microbiol. 75, 7537–7541.
Schmidt, T.S.B., Matias Rodrigues, J.F., and von Mering, C. (2014). Ecological consistency of SSU rRNA-
based operational taxonomic units at a global scale. PLoS Comput. Biol. 10, e1003594.
Schmidt, T.S.B., Matias Rodrigues, J.F., and von Mering, C. (2015). Limits to Robustness and Reproducibility
in the Demarcation of Operational Taxonomic Units. Environ. Microbiol. 17, 1689–1706.
Schommer, N.N., and Gallo, R.L. (2013). Structure and function of the human skin microbiome. Trends
Microbiol. 21, 660–668.
Schrezenmeir, J., and de Vrese, M. (2001). Probiotics, prebiotics, and synbiotics--approaching a definition.
Am. J. Clin. Nutr. 73, 361S – 364S.
Schwarz, R., Le Roux, J., Schaller, R., and Neurand, K. (1979). Micromorphology of the skin (epidermis,
dermis, subcutis) of the dog. Onderstepoort J. Vet. Res. 46, 105–109.
Scott, D.W., Miller, W.H. (William H., and Griffin, C.E. (1995). Muller and Kirk’s Small Animal Dermatology
(W.B. Saunder Co).
Seekatz, A.M., Rao, K., Santhosh, K., and Young, V.B. (2016). Dynamics of the fecal microbiome in patients
with recurrent and nonrecurrent Clostridium difficile infection. Genome Med. 8, 47.
Segata, N., Izard, J., Waldron, L., Gevers, D., Miropolsky, L., Garrett, W.S., and Huttenhower, C. (2011).
Metagenomic biomarker discovery and explanation. Genome Biol. 12, R60.
Seite, S., and Bieber, T. (2015). Barrier function and microbiotic dysbiosis in atopic dermatitis. Clin. Cosmet.
Investig. Dermatol. 8, 479–483.
Sender, R., Fuchs, S., and Milo, R. (2016). Revised Estimates for the Number of Human and Bacteria Cells in
the Body. PLoS Biol. 14, e1002533.
Sergeant, M.J., Constantinidou, C., Cogan, T., Penn, C.W., and Pallen, M.J. (2012). High-Throughput
Sequencing of 16S rRNA Gene Amplicons: Effects of Extraction Procedure, Primer Length and Annealing
Temperature. PLoS One 7, e38094.
169
Shannon, C.E., and Weaver, W. (1949). The mathematical theory of information. Urbana University of
Illinois Press, 623-656.
Shannon, P., Markiel, A., Ozier, O., Baliga, N.S., Wang, J.T., Ramage, D., Amin, N., Schwikowski, B., and
Ideker, T. (2003). Cytoscape: a software environment for integrated models of biomolecular interaction
networks. Genome Res. 13, 2498–2504.
Shu, M., Wang, Y., Yu, J., Kuo, S., Coda, A., Jiang, Y., Gallo, R.L., and Huang, C.M. (2013). Fermentation of
Propionibacterium acnes, a commensal bacterium in the human skin microbiome, as skin probiotics against
methicillin-resistant Staphylococcus aureus. PLoS One 8, e55380.
Siponen, M., Kauppila, J.H., Soini, Y., and Salo, T. (2012). TLR4 and TLR9 are induced in oral lichen planus.
J. Oral Pathol. Med. 41, 741–747.
Smeekens, S.P., Huttenhower, C., Riza, A., van de Veerdonk, F.L., Zeeuwen, P.L.J.M., Schalkwijk, J., van der
Meer, J.W.M., Xavier, R.J., Netea, M.G., and Gevers, D. (2013). Skin Microbiome Imbalance in Patients with
STAT1/STAT3 Defects Impairs Innate Host Defense Responses. J. Innate Immun. 6, 253–262.
Smith, A.M., Jain, M., Mulroney, L., Garalde, D.R., and Akeson, M. (2017). Reading canonical and modified
nucleotides in 16S ribosomal RNA using nanopore direct RNA sequencing. bioRxiv.
Smith, K., and Thiboutot, D. (2008). Thematic review series: skin lipids. Sebaceous gland lipids: friend or foe?
J. Lipid Res. 49, 271–281.
Song, S.J., Lauber, C., Costello, E.K., Lozupone, C.A., Humphrey, G., Berg-Lyons, D., Caporaso, J.G.,
Knights, D., Clemente, J.C., Nakielny, S., et al. (2013). Cohabiting family members share microbiota with one
another and with their dogs. Elife 2, e00458.
Song, X., Jin, P., Qin, S., Chen, L., and Ma, F. (2012). The evolution and origin of animal Toll-like receptor
signaling pathway revealed by network-level molecular evolutionary analyses. PLoS One 7, e51657.
Song, Y., Garg, S., Girotra, M., Maddox, C., von Rosenvinge, E.C., Dutta, A., Dutta, S., and Fricke, W.F.
(2013b). Microbiota Dynamics in Patients Treated with Fecal Microbiota Transplantation for Recurrent
Clostridium difficile Infection. PLoS One 8, e81330.
Stackebrandt, E., and Ebers, J. (2006). Taxonomic parameters revisited: tarnished gold standards. Microbiol
Today 33, 152–155.
Staley, C., Vaughn, B.P., Graiziger, C.T., Singroy, S., Hamilton, M.J., Yao, D., Chen, C., Khoruts, A., and
Sadowsky, M.J. (2017). Community dynamics drive punctuated engraftment of the fecal microbiome
following transplantation using freeze-dried, encapsulated fecal microbiota. Gut Microbes 28, 1–13.
Suter, M.M., Schulze, K., Bergman, W., Welle, M., Roosje, P., and Müller, E.J. (2009). The keratinocyte in
epidermal renewal and defence. Vet. Dermatol. 20, 515–532.
Torres, S., Clayton, J.B., Danzeisen, J.L., Ward, T., Huang, H., Knights, D., and Johnson, T.J. (2017). Diverse
bacterial communities exist on canine skin and are impacted by cohabitation and time. PeerJ 5, e3075.
Travers, K.J., Chin, C.-S., Rank, D.R., Eid, J.S., and Turner, S.W. (2010). A flexible and efficient template
format for circular consensus sequencing and SNP detection. Nucleic Acids Res. 38, e159.
Tuon, F.F., Fernandes, E.R., Duarte, M.I.S., and Amato, V.S. (2010). The expression of TLR2, TLR4 and
TLR9 in the epidermis of patients with cutaneous leishmaniasis. J. Dermatol. Sci. 59, 55–57.
Ueta, M., Sotozono, C., Inatomi, T., and Kojima, K. (2007). Toll-like receptor 3 gene polymorphisms in
Japanese patients with Stevens–Johnson syndrome. Br J Ophthalmol. 91, 962-5
170
Ushijima, T., Takahashi, M., and Ozaki, Y. (1984). Acetic, propionic, and oleic acid as the possible factors
influencing the predominant residence of some species of Propionibacterium and coagulase-negative
Staphylococcus on normal human skin. Can. J. Microbiol. 30, 647–652.
van Dijk, E.L., Auger, H., Jaszczyszyn, Y., and Thermes, C. (2014). Ten years of next-generation sequencing
technology. Trends Genet. 30, 418–426.
Vázquez-Baeza, Y., Hyde, E.R., Suchodolski, J.S., and Knight, R. (2016). Dog and human inflammatory
bowel disease rely on overlapping yet distinct dysbiosis networks. Nat. Microbiol. 1, 16177.
Větrovský, T., and Baldrian, P. (2013). The Variability of the 16S rRNA Gene in Bacterial Genomes and Its
Consequences for Bacterial Community Analyses. PLoS One. 8, e57923.
Viljanen, M., Kuitunen, M., Haahtela, T., Juntunen-Backman, K., Korpela, R., and Savilahti, E. (2005).
Probiotic effects on faecal inflammatory markers and on faecal IgA in food allergic atopic eczema/dermatitis
syndrome infants. Pediatr. Allergy Immunol. 16, 65–71.
Vincent, A.T., Derome, N., Boyle, B., Culley, A.I., and Charette, S.J. (2016). Next-generation sequencing
(NGS) in the microbiological world: How to make the most of your money. J. Microbiol. Methods. S0167-
7012, 30031-8.
Virgin, H.W. (2014). The virome in mammalian physiology and disease. Cell 157, 142–150.
Volz, T., Skabytska, Y., Guenova, E., Chen, K.-M., Frick, J.-S., Kirschning, C.J., Kaesler, S., Röcken, M., and
Biedermann, T. (2014). Nonpathogenic Bacteria Alleviating Atopic Dermatitis Inflammation Induce IL-10-
Producing Dendritic Cells and Regulatory Tr1 Cells. J. Invest. Dermatol. 134, 96–104.
Wagner Mackenzie, B., Waite, D.W., and Taylor, M.W. (2015). Evaluating variation in human gut microbiota
profiles due to DNA extraction method and inter-subject differences. Front. Microbiol. 6, 130.
Walker, A.W., Martin, J.C., Scott, P., Parkhill, J., Flint, H.J., and Scott, K.P. (2015). 16S rRNA gene-based
profiling of the human infant gut microbiota is strongly influenced by sample processing and PCR primer
choice. Microbiome 3, 26.
Walters, W.A., Caporaso, J.G., Lauber, C.L., Berg-Lyons, D., Fierer, N., and Knight, R. (2011).
PrimerProspector: de novo design and taxonomic analysis of barcoded polymerase chain reaction primers.
Bioinformatics 27, 1159–1161.
Wang, Q., Garrity, G.M., Tiedje, J.M., and Cole, J.R. (2007). Naive Bayesian classifier for rapid assignment of
rRNA sequences into the new bacterial taxonomy. Appl. Environ. Microbiol. 73, 5261–5267.
Wang, Y., Yang, Q., and Wang, Z. (2015). The evolution of nanopore sequencing. Front. Genet. 5, 449.
Wanke, I., Steffen, H., Christ, C., Krismer, B., Götz, F., Peschel, A., Schaller, M., and Schittek, B. (2011). Skin
Commensals Amplify the Innate Immune Response to Pathogens by Activation of Distinct Signaling
Pathways. J. Invest. Dermatol. 131, 382–390.
Werling, D., and Jungi, T.W. (2003). TOLL-like receptors linking innate and adaptive immune response. Vet.
Immunol. Immunopathol. 91, 1–12.
Werling, D., Jann, O.C., Offord, V., Glass, E.J., and Coffey, T.J. (2009). Variation matters: TLR structure and
species-specific pathogen recognition. Trends Immunol. 30, 124–130.
Wesolowska-Andersen, A., Bahl, M., Carvalho, V., Kristiansen, K., Sicheritz-Pontén, T., Gupta, R., and Licht,
T. (2014). Choice of bacterial DNA extraction method from fecal material influences community structure as
evaluated by metagenomic analysis. Microbiome 2, 19.
171
Westcott, S.L., and Schloss, P.D. (2017). OptiClust, an Improved Method for Assigning Amplicon-Based
Sequence Data to Operational Taxonomic Units. mSphere 2, e00073-17.
White, S.D., and Yager, J.A. (1995). Resident dendritic cells in the epidermis: Langerhans cells, Merkel cells
and melanocytes. Vet. Dermatol. 6, 1–8.
Wlasiuk, G., and Nachman, M.W. (2010). Adaptation and constraint at Toll-like receptors in primates. Mol.
Biol. Evol. 27, 2172–2186.
Wolcott, R.D., Gontcharova, V., Sun, Y., and Dowd, S.E. (2009). Evaluation of the bacterial diversity among
and within individual venous leg ulcers using bacterial tag-encoded FLX and titanium amplicon
pyrosequencing and metagenomic approaches. BMC Microbiol. 9, 226.
Wong, S.H., Gochhait, S., Malhotra, D., Pettersson, F.H., Teo, Y.Y., Khor, C.C., Rautanen, A., Chapman,
S.J., Mills, T.C., Srivastava, A., et al. (2010). Leprosy and the Adaptation of Human Toll-Like Receptor 1.
PLoS Pathog. 6, e1000979.
Wylie, K.M., Truty, R.M., Sharpton, T.J., Mihindukulasuriya, K. a, Zhou, Y., Gao, H., Sodergren, E.,
Weinstock, G.M., and Pollard, K.S. (2012). Novel bacterial taxa in the human microbiome. PLoS One 7,
e35294.
Yamasaki, K., Kanada, K., Macleod, D.T., Borkowski, A.W., Morizane, S., Nakatsuji, T., Cogen, A.L., and
Gallo, R.L. (2011). TLR2 Expression Is Increased in Rosacea and Stimulates Enhanced Serine Protease
Production by Keratinocytes. J. Invest. Dermatol. 131, 688–697.
Yarza, P., Yilmaz, P., Pruesse, E., Glöckner, F.O., Ludwig, W., Schleifer, K.-H., Whitman, W.B., Euzéby, J.,
Amann, R., and Rosselló-Móra, R. (2014). Uniting the classification of cultured and uncultured bacteria and
archaea using 16S rRNA gene sequences. Nat. Rev. Microbiol. 12, 635–645.
Ying, S., Zeng, D., Chi, L., Tan, Y., Galzote, C., and Cardona, C. (2015). The Influence of Age and Gender
on Skin- Associated Microbial Communities in Urban and Rural Human Populations. PLoS One. 10,
e0141842
Yokoyama, N., Ohta, H., Yamazaki, J., Kagawa, Y., Ichii, O., Khoirun, N., Morita, T., Osuga, T., Lim, S.Y.,
Sasaki, N., et al. (2017). Localization of Toll-like Receptor (TLR) 2 and TLR4 mRNA in the Colorectal
Mucosa of Miniature Dachshunds with Inflammatory Colorectal Polyps. J. Comp. Pathol. 156, 183–190.
Yuan, S., Cohen, D.B., Ravel, J., Abdo, Z., and Forney, L.J. (2012). Evaluation of Methods for the Extraction
and Purification of DNA from the Human Microbiome. PLoS One 7, e33865.
Zahl, S. (1977). Jackknifing An Index of Diversity. Ecology 58, 907–913.
Zeeuwen, P.L., Boekhorst, J., van den Bogaard, E.H., de Koning, H.D., van de Kerkhof, P.M., Saulnier,
D.M., van Swam, I.I., van Hijum, S.A., Kleerebezem, M., Schalkwijk, J., et al. (2012). Microbiome dynamics
of human epidermis following skin barrier disruption. Genome Biol. 13, R101.
Zhang, E., Tanaka, T., Tajima, M., Tsuboi, R., Nishikawa, A., and Sugita, T. (2011). Characterization of the
skin fungal microbiota in patients with atopic dermatitis and in healthy subjects. Microbiol. Immunol. 55,
625–632.
Zhou, H., Gu, Æ.J., and Lamont, Æ.S.J. (2007). Evolutionary Analysis for Functional Divergence of the Toll-
Like Receptor Gene Family and Altered Functional Constraints. 65, 119–123.
top related